First Post

Artificial Intelligence Nanodegree¶

Computer Vision Capstone¶

Project: Facial Keypoint Detection¶

Welcome to the final Computer Vision project in the Artificial Intelligence Nanodegree program!

In this project, you’ll combine your knowledge of computer vision techniques and deep learning to build and end-to-end facial keypoint recognition system! Facial keypoints include points around the eyes, nose, and mouth on any face and are used in many applications, from facial tracking to emotion recognition.

There are three main parts to this project:

Part 1 : Investigating OpenCV, pre-processing, and face detection

Part 2 : Training a Convolutional Neural Network (CNN) to detect facial keypoints

Part 3 : Putting parts 1 and 2 together to identify facial keypoints on any image!

*Here's what you need to know to complete the project:

In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested.

a. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!

In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation.

a. Each section where you will answer a question is preceded by a 'Question X' header.

b. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'.

Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.

The rubric contains optional suggestions for enhancing the project beyond the minimum requirements. If you decide to pursue the "(Optional)" sections, you should include the code in this IPython notebook.

Your project submission will be evaluated based on your answers to each of the questions and the code implementations you provide.

Steps to Complete the Project¶

Each part of the notebook is further broken down into separate steps. Feel free to use the links below to navigate the notebook.

In this project you will get to explore a few of the many computer vision algorithms built into the OpenCV library. This expansive computer vision library is now almost 20 years old and still growing!

The project itself is broken down into three large parts, then even further into separate steps. Make sure to read through each step, and complete any sections that begin with '(IMPLEMENTATION)' in the header; these implementation sections may contain multiple TODOs that will be marked in code. For convenience, we provide links to each of these steps below.

Part 1 : Investigating OpenCV, pre-processing, and face detection

Step 0: Detect Faces Using a Haar Cascade Classifier
Step 1: Add Eye Detection
Step 2: De-noise an Image for Better Face Detection
Step 3: Blur an Image and Perform Edge Detection
Step 4: Automatically Hide the Identity of an Individual

Part 2 : Training a Convolutional Neural Network (CNN) to detect facial keypoints

Step 5: Create a CNN to Recognize Facial Keypoints
Step 6: Compile and Train the Model
Step 7: Visualize the Loss and Answer Questions

Part 3 : Putting parts 1 and 2 together to identify facial keypoints on any image!

Step 8: Build a Robust Facial Keypoints Detector (Complete the CV Pipeline)

Step 0: Detect Faces Using a Haar Cascade Classifier¶

Have you ever wondered how Facebook automatically tags images with your friends' faces? Or how high-end cameras automatically find and focus on a certain person's face? Applications like these depend heavily on the machine learning task known as face detection - which is the task of automatically finding faces in images containing people.

At its root face detection is a classification problem - that is a problem of distinguishing between distinct classes of things. With face detection these distinct classes are 1) images of human faces and 2) everything else.

We use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images. OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the detector_architectures directory.

Import Resources¶

In the next python cell, we load in the required libraries for this section of the project.

In [1]:

# Import required libraries for this section

%matplotlib inline

import numpy as np
import matplotlib.pyplot as plt
import math
import cv2                     # OpenCV library for computer vision
from PIL import Image
import time

Next, we load in and display a test image for performing face detection.

Note: by default OpenCV assumes the ordering of our image's color channels are Blue, then Green, then Red. This is slightly out of order with most image types we'll use in these experiments, whose color channels are ordered Red, then Green, then Blue. In order to switch the Blue and Red channels of our test image around we will use OpenCV's cvtColor function, which you can read more about by checking out some of its documentation located here. This is a general utility function that can do other transformations too like converting a color image to grayscale, and transforming a standard color image to HSV color space.

In [2]:

# Load in color image for face detection
image = cv2.imread('images/test_image_1.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Plot our image using subplots to specify a size and title
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image)

Out[2]:

<matplotlib.image.AxesImage at 0x7fdd09360630>

There are a lot of people - and faces - in this picture. 13 faces to be exact! In the next code cell, we demonstrate how to use a Haar Cascade classifier to detect all the faces in this test image.

This face detector uses information about patterns of intensity in an image to reliably detect faces under varying light conditions. So, to use this face detector, we'll first convert the image from color to grayscale.

Then, we load in the fully trained architecture of the face detector -- found in the file haarcascade_frontalface_default.xml - and use it on our image to find faces!

To learn more about the parameters of the detector see this post.

In [3]:

# Convert the RGB  image to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)

# Extract the pre-trained face detector from an xml file
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray, 4, 6)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))

# Make a copy of the orginal image to draw face detections on
image_with_detections = np.copy(image)

# Get the bounding box for each detected face
for (x,y,w,h) in faces:
    # Add a red bounding box to the detections image
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
    

# Display the image with the detections
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Image with Face Detections')
ax1.imshow(image_with_detections)

Number of faces detected: 13

Out[3]:

<matplotlib.image.AxesImage at 0x7fdd092c23c8>

In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.

Step 1: Add Eye Detections¶

There are other pre-trained detectors available that use a Haar Cascade Classifier - including full human body detectors, license plate detectors, and more. A full list of the pre-trained architectures can be found here.

To test your eye detector, we'll first read in a new test image with just a single face.

In [4]:

# Load in color image for face detection
image = cv2.imread('images/james.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Plot the RGB image
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image)

Out[4]:

<matplotlib.image.AxesImage at 0x7fdd092e8668>

Notice that even though the image is a black and white image, we have read it in as a color image and so it will still need to be converted to grayscale in order to perform the most accurate face detection.

So, the next steps will be to convert this image to grayscale, then load OpenCV's face detector and run it with parameters that detect this face accurately.

In [5]:

# Convert the RGB  image to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)

# Extract the pre-trained face detector from an xml file
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray, 1.25, 6)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))

# Make a copy of the orginal image to draw face detections on
image_with_detections = np.copy(image)

# Get the bounding box for each detected face
for (x,y,w,h) in faces:
    # Add a red bounding box to the detections image
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
    

# Display the image with the detections
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Image with Face Detection')
ax1.imshow(image_with_detections)

Number of faces detected: 1

Out[5]:

<matplotlib.image.AxesImage at 0x7fdd09293668>

(IMPLEMENTATION) Add an eye detector to the current face detection setup.¶

A Haar-cascade eye detector can be included in the same way that the face detector was and, in this first task, it will be your job to do just this.

To set up an eye detector, use the stored parameters of the eye cascade detector, called haarcascade_eye.xml, located in the detector_architectures subdirectory. In the next code cell, create your eye detector and store its detections.

A few notes before you get started:

First, make sure to give your loaded eye detector the variable name

eye_cascade

and give the list of eye regions you detect the variable name

eyes

Second, since we've already run the face detector over this image, you should only search for eyes within the rectangular face regions detected in faces. This will minimize false detections.

Lastly, once you've run your eye detector over the facial detection region, you should display the RGB image with both the face detection boxes (in red) and your eye detections (in green) to verify that everything works as expected.

In [6]:

# Make a copy of the original image to plot rectangle detections
image_with_detections = np.copy(image)   

# Loop over the detections and draw their corresponding face detection boxes
for (x,y,w,h) in faces:
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h),(255,0,0), 3)  
    
# Do not change the code above this comment!

    
## TODO: Add eye detection, using haarcascade_eye.xml, to the current face detector algorithm
## TODO: Loop over the eye detections and draw their corresponding boxes in green on image_with_detections
eye_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_eye.xml')
for (x, y, w, h) in faces:
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h),(255,0,0), 3) 
    roi_gray = gray[y:y+h, x:x+w]
    roi_color = image_with_detections[y:y+h, x:x+w]
    eyes = eye_cascade.detectMultiScale(roi_gray, 2,1)
    print('number of eyes detected: ', len(eyes))
    for (dx, dy, dw, dh) in eyes:
        cv2.rectangle(roi_color, (dx,dy), (dx+dw, dy+dh), (0,255,0),2)
# Plot the image with both faces and eyes detected
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Image with Face and Eye Detection')
ax1.imshow(image_with_detections)

number of eyes detected:  2

Out[6]:

<matplotlib.image.AxesImage at 0x7fdd09249160>

(Optional) Add face and eye detection to your laptop camera¶

It's time to kick it up a notch, and add face and eye detection to your laptop's camera! Afterwards, you'll be able to show off your creation like in the gif shown below - made with a completed version of the code!

Notice that not all of the detections here are perfect - and your result need not be perfect either. You should spend a small amount of time tuning the parameters of your detectors to get reasonable results, but don't hold out for perfection. If we wanted perfection we'd need to spend a ton of time tuning the parameters of each detector, cleaning up the input image frames, etc. You can think of this as more of a rapid prototype.

The next cell contains code for a wrapper function called laptop_camera_face_eye_detector that, when called, will activate your laptop's camera. You will place the relevant face and eye detection code in this wrapper function to implement face/eye detection and mark those detections on each image frame that your camera captures.

Before adding anything to the function, you can run it to get an idea of how it works - a small window should pop up showing you the live feed from your camera; you can press any key to close this window.

Note: Mac users may find that activating this function kills the kernel of their notebook every once in a while. If this happens to you, just restart your notebook's kernel, activate cell(s) containing any crucial import statements, and you'll be good to go!

In [7]:

### Add face and eye detection to this laptop camera function 
# Make sure to draw out all faces/eyes found in each frame on the shown video feed

import cv2
import time 

# wrapper function for face/eye detection with your laptop camera
def laptop_camera_go():
    # Create instance of video capturer
    cv2.namedWindow("face detection activated")
    vc = cv2.VideoCapture(0)

    # Try to get the first frame
    if vc.isOpened(): 
        rval, frame = vc.read()
    else:
        rval = False
    
    # Keep the video stream open
    while rval:
        # Plot the image from camera with all the face and eye detections marked
        cv2.imshow("face detection activated", frame)
        
        # Exit functionality - press any key to exit laptop video
        key = cv2.waitKey(20)
        if key > 0: # Exit by pressing any key
            # Destroy windows 
            cv2.destroyAllWindows()
            
            # Make sure window closes on OSx
            for i in range (1,5):
                cv2.waitKey(1)
            return
        
        # Read next frame
        time.sleep(0.05)             # control framerate for computation - default 20 frames per sec
        rval, frame = vc.read()

In [ ]:

# Call the laptop camera face/eye detector function above
laptop_camera_go()

Step 2: De-noise an Image for Better Face Detection¶

Image quality is an important aspect of any computer vision task. Typically, when creating a set of images to train a deep learning network, significant care is taken to ensure that training images are free of visual noise or artifacts that hinder object detection. While computer vision algorithms - like a face detector - are typically trained on 'nice' data such as this, new test data doesn't always look so nice!

When applying a trained computer vision algorithm to a new piece of test data one often cleans it up first before feeding it in. This sort of cleaning - referred to as pre-processing - can include a number of cleaning phases like blurring, de-noising, color transformations, etc., and many of these tasks can be accomplished using OpenCV.

In this short subsection we explore OpenCV's noise-removal functionality to see how we can clean up a noisy image, which we then feed into our trained face detector.

Create a noisy image to work with¶

In the next cell, we create an artificial noisy version of the previous multi-face image. This is a little exaggerated - we don't typically get images that are this noisy - but image noise, or 'grainy-ness' in a digitial image - is a fairly common phenomenon.

In [8]:

# Load in the multi-face test image again
image = cv2.imread('images/test_image_1.jpg')

# Convert the image copy to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Make an array copy of this image
image_with_noise = np.asarray(image)

# Create noise - here we add noise sampled randomly from a Gaussian distribution: a common model for noise
noise_level = 40
noise = np.random.randn(image.shape[0],image.shape[1],image.shape[2])*noise_level

# Add this noise to the array image copy
image_with_noise = image_with_noise + noise

# Convert back to uint8 format
image_with_noise = np.asarray([np.uint8(np.clip(i,0,255)) for i in image_with_noise])

# Plot our noisy image!
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Noisy Image')
ax1.imshow(image_with_noise)

Out[8]:

<matplotlib.image.AxesImage at 0x7fdd0926ccc0>

In the context of face detection, the problem with an image like this is that - due to noise - we may miss some faces or get false detections.

In the next cell we apply the same trained OpenCV detector with the same settings as before, to see what sort of detections we get.

In [9]:

# Convert the RGB  image to grayscale
gray_noise = cv2.cvtColor(image_with_noise, cv2.COLOR_RGB2GRAY)

# Extract the pre-trained face detector from an xml file
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray_noise, 4, 6)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))

# Make a copy of the orginal image to draw face detections on
image_with_detections = np.copy(image_with_noise)

# Get the bounding box for each detected face
for (x,y,w,h) in faces:
    # Add a red bounding box to the detections image
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
    

# Display the image with the detections
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Noisy Image with Face Detections')
ax1.imshow(image_with_detections)

Number of faces detected: 12

Out[9]:

<matplotlib.image.AxesImage at 0x7fdcc3889518>

With this added noise we now miss one of the faces!

(IMPLEMENTATION) De-noise this image for better face detection¶

Time to get your hands dirty: using OpenCV's built in color image de-noising functionality called fastNlMeansDenoisingColored - de-noise this image enough so that all the faces in the image are properly detected. Once you have cleaned the image in the next cell, use the cell that follows to run our trained face detector over the cleaned image to check out its detections.

You can find its official documentation here and a useful example here.

Note: you can keep all parameters except photo_render fixed as shown in the second link above. Play around with the value of this parameter - see how it affects the resulting cleaned image.

In [10]:

## TODO: Use OpenCV's built in color image de-noising function to clean up our noisy image!

denoised_image = cv2.fastNlMeansDenoisingColored(image_with_noise, None, 15, 10, 7, 21)# your final de-noised image (should be RGB)

In [11]:

## TODO: Run the face detector on the de-noised image to improve your detections and display the result
denoised_gray = cv2.cvtColor(denoised_image, cv2.COLOR_RGB2GRAY)
faces = face_cascade.detectMultiScale(denoised_gray, 2, 3)
print('number of faces detected: ', len(faces))
denoised_copy = np.copy(denoised_image)
for (x, y, w, h) in faces:
    cv2.rectangle(denoised_copy, (x,y), (x+w, y+h), (0,255,0), 3)
fig = plt.figure(figsize=(8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('De-noised image with face detection')
ax1.imshow(denoised_copy)

number of faces detected:  13

Out[11]:

<matplotlib.image.AxesImage at 0x7fdcc38adda0>

Step 3: Blur an Image and Perform Edge Detection¶

Now that we have developed a simple pipeline for detecting faces using OpenCV - let's start playing around with a few fun things we can do with all those detected faces!

Importance of Blur in Edge Detection¶

Edge detection is a concept that pops up almost everywhere in computer vision applications, as edge-based features (as well as features built on top of edges) are often some of the best features for e.g., object detection and recognition problems.

Edge detection is a dimension reduction technique - by keeping only the edges of an image we get to throw away a lot of non-discriminating information. And typically the most useful kind of edge-detection is one that preserves only the important, global structures (ignoring local structures that aren't very discriminative). So removing local structures / retaining global structures is a crucial pre-processing step to performing edge detection in an image, and blurring can do just that.

Below is an animated gif showing the result of an edge-detected cat taken from Wikipedia, where the image is gradually blurred more and more prior to edge detection. When the animation begins you can't quite make out what it's a picture of, but as the animation evolves and local structures are removed via blurring the cat becomes visible in the edge-detected image.

Edge detection is a convolution performed on the image itself, and you can read about Canny edge detection on this OpenCV documentation page.

Canny edge detection¶

In the cell below we load in a test image, then apply Canny edge detection on it. The original image is shown on the left panel of the figure, while the edge-detected version of the image is shown on the right. Notice how the result looks very busy - there are too many little details preserved in the image before it is sent to the edge detector. When applied in computer vision applications, edge detection should preserve global structure; doing away with local structures that don't help describe what objects are in the image.

In [12]:

# Load in the image
image = cv2.imread('images/fawzia.jpg')

# Convert to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Convert to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)  

# Perform Canny edge detection
edges = cv2.Canny(gray,100,200)

# Dilate the image to amplify edges
edges = cv2.dilate(edges, None)

# Plot the RGB and edge-detected image
fig = plt.figure(figsize = (15,15))
ax1 = fig.add_subplot(121)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image)

ax2 = fig.add_subplot(122)
ax2.set_xticks([])
ax2.set_yticks([])

ax2.set_title('Canny Edges')
ax2.imshow(edges, cmap='gray')

Out[12]:

<matplotlib.image.AxesImage at 0x7fdcc37fbb00>

Without first blurring the image, and removing small, local structures, a lot of irrelevant edge content gets picked up and amplified by the detector (as shown in the right panel above).

(IMPLEMENTATION) Blur the image then perform edge detection¶

In the next cell, you will repeat this experiment - blurring the image first to remove these local structures, so that only the important boudnary details remain in the edge-detected image.

Blur the image by using OpenCV's filter2d functionality - which is discussed in this documentation page - and use an averaging kernel of width equal to 4.

In [13]:

### TODO: Blur the test imageusing OpenCV's filter2d functionality, 
# Use an averaging kernel, and a kernel width equal to 4
kernel = np.ones((4,4), np.float32)/16
dst = cv2.filter2D(image, -1, kernel)
## TODO: Then perform Canny edge detection and display the output

gray1 = cv2.cvtColor(dst, cv2.COLOR_RGB2GRAY)  

# Perform Canny edge detection
edges1 = cv2.Canny(gray1,100,200)

# Dilate the image to amplify edges
edges1 = cv2.dilate(edges1, None)

# Plot the RGB and edge-detected image
fig = plt.figure(figsize = (15,15))
ax1 = fig.add_subplot(131)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image)

ax2 = fig.add_subplot(132)
ax2.set_xticks([])
ax2.set_yticks([])

ax2.set_title('Canny Edges')
ax2.imshow(edges, cmap='gray')

ax3= fig.add_subplot(133)
ax3.set_xticks([])
ax3.set_yticks([])

ax3.set_title('Canny Edges with Blur')
ax3.imshow(edges1, cmap='gray')

Out[13]:

<matplotlib.image.AxesImage at 0x7fdcc3770470>

Step 4: Automatically Hide the Identity of an Individual¶

If you film something like a documentary or reality TV, you must get permission from every individual shown on film before you can show their face, otherwise you need to blur it out - by blurring the face a lot (so much so that even the global structures are obscured)! This is also true for projects like Google's StreetView maps - an enormous collection of mapping images taken from a fleet of Google vehicles. Because it would be impossible for Google to get the permission of every single person accidentally captured in one of these images they blur out everyone's faces, the detected images must automatically blur the identity of detected people. Here's a few examples of folks caught in the camera of a Google street view vehicle.

Read in an image to perform identity detection¶

Let's try this out for ourselves. Use the face detection pipeline built above and what you know about using the filter2D to blur and image, and use these in tandem to hide the identity of the person in the following image - loaded in and printed in the next cell.

In [14]:

# Load in the image
image = cv2.imread('images/gus.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Display the image
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.set_title('Original Image')
ax1.imshow(image)

Out[14]:

<matplotlib.image.AxesImage at 0x7fdcc37a10f0>

(IMPLEMENTATION) Use blurring to hide the identity of an individual in an image¶

The idea here is to 1) automatically detect the face in this image, and then 2) blur it out! Make sure to adjust the parameters of the averaging blur filter to completely obscure this person's identity.

In [15]:

## TODO: Implement face detection
image_copy = np.copy(image)
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')
faces = face_cascade.detectMultiScale(gray, 2, 4)
print('number of faces detected: ', len(faces))
## TODO: Blur the bounding box around each detected face using an averaging filter and display the result

for (x, y, w, h) in faces:
    #cv2.rectangle(image_copy, (x,y), (x+w, y+h), (255, 0, 0), 3)
    crop_image = image_copy[y:y+h, x:x+w]
    kernel = np.ones((60,60), np.float32)/3600
    blur_image=cv2.filter2D(crop_image, -1, kernel)
    image_copy[y:y+h, x:x+w]=blur_image

fig=plt.figure(figsize=(8,8))
ax1=fig.add_subplot(111)
ax1.imshow(image_copy)
ax1.set_xticks([])
ax1.set_yticks([])
plt.show()

number of faces detected:  1

(Optional) Build identity protection into your laptop camera¶

In this optional task you can add identity protection to your laptop camera, using the previously completed code where you added face detection to your laptop camera - and the task above. You should be able to get reasonable results with little parameter tuning - like the one shown in the gif below.

As with the previous video task, to make this perfect would require significant effort - so don't strive for perfection here, strive for reasonable quality.

The next cell contains code a wrapper function called laptop_camera_identity_hider that - when called - will activate your laptop's camera. You need to place the relevant face detection and blurring code developed above in this function in order to blur faces entering your laptop camera's field of view.

Before adding anything to the function you can call it to get a hang of how it works - a small window will pop up showing you the live feed from your camera, you can press any key to close this window.

In [ ]:

### Insert face detection and blurring code into the wrapper below to create an identity protector on your laptop!
import cv2
import time 

def laptop_camera_go():
    # Create instance of video capturer
    cv2.namedWindow("face detection activated")
    vc = cv2.VideoCapture(0)

    # Try to get the first frame
    if vc.isOpened(): 
        rval, frame = vc.read()
    else:
        rval = False
    
    # Keep video stream open
    while rval:
        # Plot image from camera with detections marked
        cv2.imshow("face detection activated", frame)
        
        # Exit functionality - press any key to exit laptop video
        key = cv2.waitKey(20)
        if key > 0: # Exit by pressing any key
            # Destroy windows
            cv2.destroyAllWindows()
            
            for i in range (1,5):
                cv2.waitKey(1)
            return
        
        # Read next frame
        time.sleep(0.05)             # control framerate for computation - default 20 frames per sec
        rval, frame = vc.read()

In [ ]:

# Run laptop identity hider
laptop_camera_go()

Step 5: Create a CNN to Recognize Facial Keypoints¶

OpenCV is often used in practice with other machine learning and deep learning libraries to produce interesting results. In this stage of the project you will create your own end-to-end pipeline - employing convolutional networks in keras along with OpenCV - to apply a "selfie" filter to streaming video and images.

You will start by creating and then training a convolutional network that can detect facial keypoints in a small dataset of cropped images of human faces. We then guide you towards OpenCV to expanding your detection algorithm to more general images. What are facial keypoints? Let's take a look at some examples.

Facial keypoints (also called facial landmarks) are the small blue-green dots shown on each of the faces in the image above - there are 15 keypoints marked in each image. They mark important areas of the face - the eyes, corners of the mouth, the nose, etc. Facial keypoints can be used in a variety of machine learning applications from face and emotion recognition to commercial applications like the image filters popularized by Snapchat.

Below we illustrate a filter that, using the results of this section, automatically places sunglasses on people in images (using the facial keypoints to place the glasses correctly on each face). Here, the facial keypoints have been colored lime green for visualization purposes.

Make a facial keypoint detector¶

But first things first: how can we make a facial keypoint detector? Well, at a high level, notice that facial keypoint detection is a regression problem. A single face corresponds to a set of 15 facial keypoints (a set of 15 corresponding $(x, y)$ coordinates, i.e., an output point). Because our input data are images, we can employ a convolutional neural network to recognize patterns in our images and learn how to identify these keypoint given sets of labeled data.

In order to train a regressor, we need a training set - a set of facial image / facial keypoint pairs to train on. For this we will be using this dataset from Kaggle. We've already downloaded this data and placed it in the data directory. Make sure that you have both the training and test data files. The training dataset contains several thousand $96 \times 96$ grayscale images of cropped human faces, along with each face's 15 corresponding facial keypoints (also called landmarks) that have been placed by hand, and recorded in $(x, y)$ coordinates. This wonderful resource also has a substantial testing set, which we will use in tinkering with our convolutional network.

To load in this data, run the Python cell below - notice we will load in both the training and testing sets.

The load_data function is in the included utils.py file.

In [18]:

from utils import *

# Load training set
X_train, y_train = load_data()
print("X_train.shape == {}".format(X_train.shape))
print("y_train.shape == {}; y_train.min == {:.3f}; y_train.max == {:.3f}".format(
    y_train.shape, y_train.min(), y_train.max()))

# Load testing set
X_test, _ = load_data(test=True)
print("X_test.shape == {}".format(X_test.shape))

X_train.shape == (2140, 96, 96, 1)
y_train.shape == (2140, 30); y_train.min == -0.920; y_train.max == 0.996
X_test.shape == (1783, 96, 96, 1)

The load_data function in utils.py originates from this excellent blog post, which you are strongly encouraged to read. Please take the time now to review this function. Note how the output values - that is, the coordinates of each set of facial landmarks - have been normalized to take on values in the range $[-1, 1]$, while the pixel values of each input point (a facial image) have been normalized to the range $[0,1]$.

Note: the original Kaggle dataset contains some images with several missing keypoints. For simplicity, the load_data function removes those images with missing labels from the dataset. As an optional extension, you are welcome to amend the load_data function to include the incomplete data points.

Visualize the Training Data¶

Execute the code cell below to visualize a subset of the training data.

In [20]:

import matplotlib.pyplot as plt
%matplotlib inline

fig = plt.figure(figsize=(20,20))
fig.subplots_adjust(left=0, right=1, bottom=0, top=1, hspace=0.05, wspace=0.05)
for i in range(9):
    ax = fig.add_subplot(3, 3, i + 1, xticks=[], yticks=[])
    plot_data(X_train[i], y_train[i], ax)

For each training image, there are two landmarks per eyebrow (four total), three per eye (six total), four for the mouth, and one for the tip of the nose.

Review the plot_data function in utils.py to understand how the 30-dimensional training labels in y_train are mapped to facial locations, as this function will prove useful for your pipeline.

(IMPLEMENTATION) Specify the CNN Architecture¶

In this section, you will specify a neural network for predicting the locations of facial keypoints. Use the code cell below to specify the architecture of your neural network. We have imported some layers that you may find useful for this task, but if you need to use more Keras layers, feel free to import them in the cell.

Your network should accept a $96 \times 96$ grayscale image as input, and it should output a vector with 30 entries, corresponding to the predicted (horizontal and vertical) locations of 15 facial keypoints. If you are not sure where to start, you can find some useful starting architectures in this blog, but you are not permitted to copy any of the architectures that you find online.

In [16]:

# Import deep learning resources from Keras
from keras.models import Sequential
from keras.layers import Convolution2D, MaxPooling2D, Dropout, Activation
from keras.layers import Flatten, Dense
from keras.optimizers import SGD, RMSprop, Adagrad, Adadelta, Adam, Adamax, Nadam
from keras.callbacks import ModelCheckpoint

## TODO: Specify a CNN architecture
# Your model should accept 96x96 pixel graysale images in
# It should have a fully-connected output layer with 30 values (2 for each facial keypoint)

model = Sequential()

model.add(Convolution2D(8, kernel_size=5, padding='same',input_shape=X_train.shape[1:]))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Convolution2D(16, kernel_size=3, padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Flatten())
model.add(Dense(30))
# Summarize the model
model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_1 (Conv2D)            (None, 96, 96, 8)         208       
_________________________________________________________________
activation_1 (Activation)    (None, 96, 96, 8)         0         
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 24, 24, 8)         0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 24, 24, 16)        1168      
_________________________________________________________________
activation_2 (Activation)    (None, 24, 24, 16)        0         
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 6, 6, 16)          0         
_________________________________________________________________
flatten_1 (Flatten)          (None, 576)               0         
_________________________________________________________________
dense_1 (Dense)              (None, 30)                17310     
=================================================================
Total params: 18,686
Trainable params: 18,686
Non-trainable params: 0
_________________________________________________________________

Step 6: Compile and Train the Model¶

After specifying your architecture, you'll need to compile and train the model to detect facial keypoints'

(IMPLEMENTATION) Compile and Train the Model¶

Use the compile method to configure the learning process. Experiment with your choice of optimizer; you may have some ideas about which will work best (SGD vs. RMSprop, etc), but take the time to empirically verify your theories.

Use the fit method to train the model. Break off a validation set by setting validation_split=0.2. Save the returned History object in the history variable.

Experiment with your model to minimize the validation loss (measured as mean squared error). A very good model will achieve about 0.0015 loss (though it's possible to do even better). When you have finished training, save your model as an HDF5 file with file path my_model.h5.

SGD¶

In [17]:

from keras.optimizers import SGD, RMSprop, Adagrad, Adadelta, Adam, Adamax, Nadam
from keras.callbacks import ModelCheckpoint
## TODO: Compile the model
model.compile( optimizer='SGD', loss='mean_squared_error', metrics=['accuracy'])

checkpointer=ModelCheckpoint(save_best_only=True, filepath='sgd_model.h5')
## TODO: Train the model
SGD_hist = model.fit(X_train, y_train, batch_size=128, epochs=100, verbose=1, validation_split=0.2, callbacks=[checkpointer])

## TODO: Save the model as model.h5
model.save('sgd_model.h5')

Train on 1712 samples, validate on 428 samples
Epoch 1/100
1712/1712 [==============================] - 1s - loss: 0.1829 - acc: 0.0000e+00 - val_loss: 0.1503 - val_acc: 0.0000e+00
Epoch 2/100
1712/1712 [==============================] - 0s - loss: 0.1316 - acc: 5.8411e-04 - val_loss: 0.1167 - val_acc: 0.0000e+00
Epoch 3/100
1712/1712 [==============================] - 0s - loss: 0.1033 - acc: 0.0012 - val_loss: 0.0911 - val_acc: 0.0093
Epoch 4/100
1712/1712 [==============================] - 0s - loss: 0.0796 - acc: 0.0222 - val_loss: 0.0679 - val_acc: 0.0631
Epoch 5/100
1712/1712 [==============================] - 0s - loss: 0.0583 - acc: 0.1268 - val_loss: 0.0481 - val_acc: 0.2103
Epoch 6/100
1712/1712 [==============================] - 0s - loss: 0.0412 - acc: 0.3026 - val_loss: 0.0336 - val_acc: 0.3715
Epoch 7/100
1712/1712 [==============================] - 0s - loss: 0.0295 - acc: 0.4340 - val_loss: 0.0244 - val_acc: 0.4603
Epoch 8/100
1712/1712 [==============================] - 0s - loss: 0.0225 - acc: 0.5164 - val_loss: 0.0193 - val_acc: 0.5164
Epoch 9/100
1712/1712 [==============================] - 0s - loss: 0.0187 - acc: 0.5713 - val_loss: 0.0166 - val_acc: 0.5724
Epoch 10/100
1712/1712 [==============================] - 0s - loss: 0.0166 - acc: 0.5975 - val_loss: 0.0152 - val_acc: 0.5818
Epoch 11/100
1712/1712 [==============================] - 0s - loss: 0.0155 - acc: 0.6011 - val_loss: 0.0144 - val_acc: 0.5818
Epoch 12/100
1712/1712 [==============================] - 0s - loss: 0.0149 - acc: 0.6016 - val_loss: 0.0139 - val_acc: 0.5935
Epoch 13/100
1712/1712 [==============================] - 0s - loss: 0.0145 - acc: 0.5975 - val_loss: 0.0136 - val_acc: 0.5841
Epoch 14/100
1712/1712 [==============================] - 0s - loss: 0.0142 - acc: 0.5946 - val_loss: 0.0134 - val_acc: 0.5841
Epoch 15/100
1712/1712 [==============================] - 0s - loss: 0.0139 - acc: 0.5940 - val_loss: 0.0131 - val_acc: 0.5818
Epoch 16/100
1712/1712 [==============================] - 0s - loss: 0.0137 - acc: 0.5935 - val_loss: 0.0130 - val_acc: 0.5771
Epoch 17/100
1712/1712 [==============================] - 0s - loss: 0.0135 - acc: 0.5935 - val_loss: 0.0128 - val_acc: 0.5771
Epoch 18/100
1712/1712 [==============================] - 0s - loss: 0.0133 - acc: 0.5958 - val_loss: 0.0126 - val_acc: 0.5771
Epoch 19/100
1712/1712 [==============================] - 0s - loss: 0.0132 - acc: 0.5987 - val_loss: 0.0125 - val_acc: 0.5794
Epoch 20/100
1712/1712 [==============================] - 0s - loss: 0.0130 - acc: 0.5993 - val_loss: 0.0123 - val_acc: 0.5771
Epoch 21/100
1712/1712 [==============================] - 0s - loss: 0.0128 - acc: 0.6011 - val_loss: 0.0122 - val_acc: 0.5818
Epoch 22/100
1712/1712 [==============================] - 0s - loss: 0.0127 - acc: 0.6051 - val_loss: 0.0120 - val_acc: 0.5794
Epoch 23/100
1712/1712 [==============================] - 0s - loss: 0.0125 - acc: 0.6046 - val_loss: 0.0119 - val_acc: 0.5841
Epoch 24/100
1712/1712 [==============================] - 0s - loss: 0.0124 - acc: 0.6098 - val_loss: 0.0117 - val_acc: 0.5888
Epoch 25/100
1712/1712 [==============================] - 0s - loss: 0.0123 - acc: 0.6116 - val_loss: 0.0116 - val_acc: 0.5935
Epoch 26/100
1712/1712 [==============================] - 0s - loss: 0.0121 - acc: 0.6145 - val_loss: 0.0115 - val_acc: 0.5911
Epoch 27/100
1712/1712 [==============================] - 0s - loss: 0.0120 - acc: 0.6168 - val_loss: 0.0114 - val_acc: 0.5935
Epoch 28/100
1712/1712 [==============================] - 0s - loss: 0.0118 - acc: 0.6186 - val_loss: 0.0112 - val_acc: 0.5935
Epoch 29/100
1712/1712 [==============================] - 0s - loss: 0.0117 - acc: 0.6197 - val_loss: 0.0111 - val_acc: 0.5981
Epoch 30/100
1712/1712 [==============================] - 0s - loss: 0.0116 - acc: 0.6209 - val_loss: 0.0110 - val_acc: 0.6028
Epoch 31/100
1712/1712 [==============================] - 0s - loss: 0.0115 - acc: 0.6221 - val_loss: 0.0109 - val_acc: 0.6051
Epoch 32/100
1712/1712 [==============================] - 0s - loss: 0.0114 - acc: 0.6250 - val_loss: 0.0108 - val_acc: 0.6075
Epoch 33/100
1712/1712 [==============================] - 0s - loss: 0.0112 - acc: 0.6244 - val_loss: 0.0107 - val_acc: 0.6098
Epoch 34/100
1712/1712 [==============================] - 0s - loss: 0.0111 - acc: 0.6285 - val_loss: 0.0106 - val_acc: 0.6121
Epoch 35/100
1712/1712 [==============================] - 0s - loss: 0.0110 - acc: 0.6326 - val_loss: 0.0105 - val_acc: 0.6168
Epoch 36/100
1712/1712 [==============================] - 0s - loss: 0.0109 - acc: 0.6384 - val_loss: 0.0104 - val_acc: 0.6145
Epoch 37/100
1712/1712 [==============================] - 0s - loss: 0.0108 - acc: 0.6402 - val_loss: 0.0102 - val_acc: 0.6215
Epoch 38/100
1712/1712 [==============================] - 0s - loss: 0.0107 - acc: 0.6425 - val_loss: 0.0102 - val_acc: 0.6215
Epoch 39/100
1712/1712 [==============================] - 0s - loss: 0.0106 - acc: 0.6449 - val_loss: 0.0101 - val_acc: 0.6238
Epoch 40/100
1712/1712 [==============================] - 0s - loss: 0.0105 - acc: 0.6484 - val_loss: 0.0100 - val_acc: 0.6215
Epoch 41/100
1712/1712 [==============================] - 0s - loss: 0.0104 - acc: 0.6484 - val_loss: 0.0099 - val_acc: 0.6238
Epoch 42/100
1712/1712 [==============================] - 0s - loss: 0.0103 - acc: 0.6472 - val_loss: 0.0098 - val_acc: 0.6238
Epoch 43/100
1712/1712 [==============================] - 0s - loss: 0.0102 - acc: 0.6489 - val_loss: 0.0097 - val_acc: 0.6262
Epoch 44/100
1712/1712 [==============================] - 0s - loss: 0.0101 - acc: 0.6489 - val_loss: 0.0096 - val_acc: 0.6238
Epoch 45/100
1712/1712 [==============================] - 0s - loss: 0.0100 - acc: 0.6484 - val_loss: 0.0095 - val_acc: 0.6238
Epoch 46/100
1712/1712 [==============================] - 0s - loss: 0.0099 - acc: 0.6489 - val_loss: 0.0094 - val_acc: 0.6308
Epoch 47/100
1712/1712 [==============================] - 0s - loss: 0.0098 - acc: 0.6507 - val_loss: 0.0093 - val_acc: 0.6355
Epoch 48/100
1712/1712 [==============================] - 0s - loss: 0.0097 - acc: 0.6525 - val_loss: 0.0092 - val_acc: 0.6379
Epoch 49/100
1712/1712 [==============================] - 0s - loss: 0.0096 - acc: 0.6519 - val_loss: 0.0092 - val_acc: 0.6402
Epoch 50/100
1712/1712 [==============================] - 0s - loss: 0.0095 - acc: 0.6571 - val_loss: 0.0091 - val_acc: 0.6425
Epoch 51/100
1712/1712 [==============================] - 0s - loss: 0.0094 - acc: 0.6565 - val_loss: 0.0090 - val_acc: 0.6449
Epoch 52/100
1712/1712 [==============================] - 0s - loss: 0.0094 - acc: 0.6589 - val_loss: 0.0089 - val_acc: 0.6472
Epoch 53/100
1712/1712 [==============================] - 0s - loss: 0.0093 - acc: 0.6589 - val_loss: 0.0088 - val_acc: 0.6519
Epoch 54/100
1712/1712 [==============================] - 0s - loss: 0.0092 - acc: 0.6595 - val_loss: 0.0088 - val_acc: 0.6542
Epoch 55/100
1712/1712 [==============================] - 0s - loss: 0.0091 - acc: 0.6600 - val_loss: 0.0087 - val_acc: 0.6519
Epoch 56/100
1712/1712 [==============================] - 0s - loss: 0.0090 - acc: 0.6636 - val_loss: 0.0086 - val_acc: 0.6519
Epoch 57/100
1712/1712 [==============================] - 0s - loss: 0.0090 - acc: 0.6630 - val_loss: 0.0086 - val_acc: 0.6519
Epoch 58/100
1712/1712 [==============================] - 0s - loss: 0.0089 - acc: 0.6653 - val_loss: 0.0085 - val_acc: 0.6519
Epoch 59/100
1712/1712 [==============================] - 0s - loss: 0.0088 - acc: 0.6641 - val_loss: 0.0084 - val_acc: 0.6519
Epoch 60/100
1712/1712 [==============================] - 0s - loss: 0.0087 - acc: 0.6653 - val_loss: 0.0083 - val_acc: 0.6542
Epoch 61/100
1712/1712 [==============================] - 0s - loss: 0.0087 - acc: 0.6671 - val_loss: 0.0083 - val_acc: 0.6542
Epoch 62/100
1712/1712 [==============================] - 0s - loss: 0.0086 - acc: 0.6671 - val_loss: 0.0082 - val_acc: 0.6542
Epoch 63/100
1712/1712 [==============================] - 0s - loss: 0.0085 - acc: 0.6688 - val_loss: 0.0082 - val_acc: 0.6542
Epoch 64/100
1712/1712 [==============================] - 0s - loss: 0.0085 - acc: 0.6717 - val_loss: 0.0081 - val_acc: 0.6542
Epoch 65/100
1712/1712 [==============================] - 0s - loss: 0.0084 - acc: 0.6723 - val_loss: 0.0080 - val_acc: 0.6542
Epoch 66/100
1712/1712 [==============================] - 0s - loss: 0.0083 - acc: 0.6729 - val_loss: 0.0080 - val_acc: 0.6495
Epoch 67/100
1712/1712 [==============================] - 0s - loss: 0.0083 - acc: 0.6746 - val_loss: 0.0079 - val_acc: 0.6542
Epoch 68/100
1712/1712 [==============================] - 0s - loss: 0.0082 - acc: 0.6746 - val_loss: 0.0079 - val_acc: 0.6542
Epoch 69/100
1712/1712 [==============================] - 0s - loss: 0.0081 - acc: 0.6758 - val_loss: 0.0078 - val_acc: 0.6565
Epoch 70/100
1712/1712 [==============================] - 0s - loss: 0.0081 - acc: 0.6770 - val_loss: 0.0077 - val_acc: 0.6565
Epoch 71/100
1712/1712 [==============================] - 0s - loss: 0.0080 - acc: 0.6776 - val_loss: 0.0077 - val_acc: 0.6565
Epoch 72/100
1712/1712 [==============================] - 0s - loss: 0.0080 - acc: 0.6782 - val_loss: 0.0076 - val_acc: 0.6565
Epoch 73/100
1712/1712 [==============================] - 0s - loss: 0.0079 - acc: 0.6776 - val_loss: 0.0076 - val_acc: 0.6565
Epoch 74/100
1712/1712 [==============================] - 0s - loss: 0.0079 - acc: 0.6787 - val_loss: 0.0075 - val_acc: 0.6565
Epoch 75/100
1712/1712 [==============================] - 0s - loss: 0.0078 - acc: 0.6787 - val_loss: 0.0075 - val_acc: 0.6612
Epoch 76/100
1712/1712 [==============================] - 0s - loss: 0.0077 - acc: 0.6782 - val_loss: 0.0074 - val_acc: 0.6636
Epoch 77/100
1712/1712 [==============================] - 0s - loss: 0.0077 - acc: 0.6805 - val_loss: 0.0074 - val_acc: 0.6636
Epoch 78/100
1712/1712 [==============================] - 0s - loss: 0.0076 - acc: 0.6828 - val_loss: 0.0073 - val_acc: 0.6659
Epoch 79/100
1712/1712 [==============================] - 0s - loss: 0.0076 - acc: 0.6840 - val_loss: 0.0073 - val_acc: 0.6659
Epoch 80/100
1712/1712 [==============================] - 0s - loss: 0.0075 - acc: 0.6846 - val_loss: 0.0072 - val_acc: 0.6659
Epoch 81/100
1712/1712 [==============================] - 0s - loss: 0.0075 - acc: 0.6863 - val_loss: 0.0072 - val_acc: 0.6659
Epoch 82/100
1712/1712 [==============================] - 0s - loss: 0.0074 - acc: 0.6887 - val_loss: 0.0072 - val_acc: 0.6682
Epoch 83/100
1712/1712 [==============================] - 0s - loss: 0.0074 - acc: 0.6881 - val_loss: 0.0071 - val_acc: 0.6682
Epoch 84/100
1712/1712 [==============================] - 0s - loss: 0.0073 - acc: 0.6898 - val_loss: 0.0071 - val_acc: 0.6706
Epoch 85/100
1712/1712 [==============================] - 0s - loss: 0.0073 - acc: 0.6904 - val_loss: 0.0070 - val_acc: 0.6729
Epoch 86/100
1712/1712 [==============================] - 0s - loss: 0.0073 - acc: 0.6904 - val_loss: 0.0070 - val_acc: 0.6752
Epoch 87/100
1712/1712 [==============================] - 0s - loss: 0.0072 - acc: 0.6916 - val_loss: 0.0070 - val_acc: 0.6752
Epoch 88/100
1712/1712 [==============================] - 0s - loss: 0.0072 - acc: 0.6916 - val_loss: 0.0069 - val_acc: 0.6776
Epoch 89/100
1712/1712 [==============================] - 0s - loss: 0.0071 - acc: 0.6910 - val_loss: 0.0069 - val_acc: 0.6776
Epoch 90/100
1712/1712 [==============================] - 0s - loss: 0.0071 - acc: 0.6910 - val_loss: 0.0068 - val_acc: 0.6776
Epoch 91/100
1712/1712 [==============================] - 0s - loss: 0.0070 - acc: 0.6916 - val_loss: 0.0068 - val_acc: 0.6752
Epoch 92/100
1712/1712 [==============================] - 0s - loss: 0.0070 - acc: 0.6910 - val_loss: 0.0068 - val_acc: 0.6752
Epoch 93/100
1712/1712 [==============================] - 0s - loss: 0.0070 - acc: 0.6910 - val_loss: 0.0067 - val_acc: 0.6752
Epoch 94/100
1712/1712 [==============================] - 0s - loss: 0.0069 - acc: 0.6910 - val_loss: 0.0067 - val_acc: 0.6752
Epoch 95/100
1712/1712 [==============================] - 0s - loss: 0.0069 - acc: 0.6910 - val_loss: 0.0067 - val_acc: 0.6799
Epoch 96/100
1712/1712 [==============================] - 0s - loss: 0.0069 - acc: 0.6910 - val_loss: 0.0066 - val_acc: 0.6822
Epoch 97/100
1712/1712 [==============================] - 0s - loss: 0.0068 - acc: 0.6910 - val_loss: 0.0066 - val_acc: 0.6822
Epoch 98/100
1712/1712 [==============================] - 0s - loss: 0.0068 - acc: 0.6910 - val_loss: 0.0066 - val_acc: 0.6869
Epoch 99/100
1712/1712 [==============================] - 0s - loss: 0.0067 - acc: 0.6898 - val_loss: 0.0065 - val_acc: 0.6869
Epoch 100/100
1712/1712 [==============================] - 0s - loss: 0.0067 - acc: 0.6898 - val_loss: 0.0065 - val_acc: 0.6869

RMSprop¶

In [18]:

#RMSprop
model = Sequential()

model.add(Convolution2D(8, kernel_size=5, padding='same',input_shape=X_train.shape[1:]))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Convolution2D(16, kernel_size=3, padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Flatten())
model.add(Dense(30))
# Summarize the model
model.summary()

model.compile( optimizer='RMSprop', loss='mean_squared_error', metrics=['accuracy'])

checkpointer=ModelCheckpoint(save_best_only=True, filepath='rmsprop_model.h5')
## TODO: Train the model
rmsprop_hist = model.fit(X_train, y_train, batch_size=128, epochs=100, verbose=1, validation_split=0.2, callbacks=[checkpointer])

## TODO: Save the model as model.h5
model.save('rmsprop_model.h5')

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_3 (Conv2D)            (None, 96, 96, 8)         208       
_________________________________________________________________
activation_3 (Activation)    (None, 96, 96, 8)         0         
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 24, 24, 8)         0         
_________________________________________________________________
conv2d_4 (Conv2D)            (None, 24, 24, 16)        1168      
_________________________________________________________________
activation_4 (Activation)    (None, 24, 24, 16)        0         
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 6, 6, 16)          0         
_________________________________________________________________
flatten_2 (Flatten)          (None, 576)               0         
_________________________________________________________________
dense_2 (Dense)              (None, 30)                17310     
=================================================================
Total params: 18,686
Trainable params: 18,686
Non-trainable params: 0
_________________________________________________________________
Train on 1712 samples, validate on 428 samples
Epoch 1/100
1712/1712 [==============================] - 0s - loss: 0.0401 - acc: 0.4054 - val_loss: 0.0104 - val_acc: 0.6262
Epoch 2/100
1712/1712 [==============================] - 0s - loss: 0.0118 - acc: 0.6484 - val_loss: 0.0095 - val_acc: 0.6893
Epoch 3/100
1712/1712 [==============================] - 0s - loss: 0.0101 - acc: 0.6752 - val_loss: 0.0105 - val_acc: 0.6963
Epoch 4/100
1712/1712 [==============================] - 0s - loss: 0.0091 - acc: 0.6910 - val_loss: 0.0108 - val_acc: 0.6963
Epoch 5/100
1712/1712 [==============================] - 0s - loss: 0.0081 - acc: 0.7009 - val_loss: 0.0068 - val_acc: 0.6776
Epoch 6/100
1712/1712 [==============================] - 0s - loss: 0.0076 - acc: 0.6939 - val_loss: 0.0070 - val_acc: 0.7009
Epoch 7/100
1712/1712 [==============================] - 0s - loss: 0.0070 - acc: 0.7009 - val_loss: 0.0082 - val_acc: 0.7009
Epoch 8/100
1712/1712 [==============================] - 0s - loss: 0.0064 - acc: 0.7009 - val_loss: 0.0059 - val_acc: 0.7033
Epoch 9/100
1712/1712 [==============================] - 0s - loss: 0.0062 - acc: 0.7039 - val_loss: 0.0055 - val_acc: 0.7056
Epoch 10/100
1712/1712 [==============================] - 0s - loss: 0.0057 - acc: 0.7009 - val_loss: 0.0057 - val_acc: 0.7056
Epoch 11/100
1712/1712 [==============================] - 0s - loss: 0.0055 - acc: 0.6986 - val_loss: 0.0048 - val_acc: 0.6916
Epoch 12/100
1712/1712 [==============================] - 0s - loss: 0.0051 - acc: 0.6980 - val_loss: 0.0052 - val_acc: 0.6939
Epoch 13/100
1712/1712 [==============================] - 0s - loss: 0.0049 - acc: 0.6974 - val_loss: 0.0044 - val_acc: 0.6682
Epoch 14/100
1712/1712 [==============================] - 0s - loss: 0.0048 - acc: 0.6933 - val_loss: 0.0047 - val_acc: 0.6986
Epoch 15/100
1712/1712 [==============================] - 0s - loss: 0.0046 - acc: 0.6916 - val_loss: 0.0039 - val_acc: 0.6893
Epoch 16/100
1712/1712 [==============================] - 0s - loss: 0.0045 - acc: 0.6881 - val_loss: 0.0039 - val_acc: 0.6986
Epoch 17/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.6951 - val_loss: 0.0041 - val_acc: 0.6986
Epoch 18/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.6928 - val_loss: 0.0035 - val_acc: 0.7033
Epoch 19/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.6893 - val_loss: 0.0037 - val_acc: 0.7033
Epoch 20/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.6963 - val_loss: 0.0038 - val_acc: 0.7033
Epoch 21/100
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7004 - val_loss: 0.0034 - val_acc: 0.7150
Epoch 22/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.6963 - val_loss: 0.0037 - val_acc: 0.6916
Epoch 23/100
1712/1712 [==============================] - 0s - loss: 0.0035 - acc: 0.6922 - val_loss: 0.0037 - val_acc: 0.7103
Epoch 24/100
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.6916 - val_loss: 0.0037 - val_acc: 0.7150
Epoch 25/100
1712/1712 [==============================] - 0s - loss: 0.0035 - acc: 0.6974 - val_loss: 0.0037 - val_acc: 0.7103
Epoch 26/100
1712/1712 [==============================] - 0s - loss: 0.0034 - acc: 0.6968 - val_loss: 0.0034 - val_acc: 0.7056
Epoch 27/100
1712/1712 [==============================] - 0s - loss: 0.0035 - acc: 0.6928 - val_loss: 0.0031 - val_acc: 0.7056
Epoch 28/100
1712/1712 [==============================] - 0s - loss: 0.0034 - acc: 0.6986 - val_loss: 0.0032 - val_acc: 0.7126
Epoch 29/100
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7033 - val_loss: 0.0032 - val_acc: 0.7196
Epoch 30/100
1712/1712 [==============================] - 0s - loss: 0.0033 - acc: 0.7015 - val_loss: 0.0035 - val_acc: 0.7150
Epoch 31/100
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7027 - val_loss: 0.0038 - val_acc: 0.7196
Epoch 32/100
1712/1712 [==============================] - 0s - loss: 0.0031 - acc: 0.7044 - val_loss: 0.0035 - val_acc: 0.7103
Epoch 33/100
1712/1712 [==============================] - 0s - loss: 0.0031 - acc: 0.7109 - val_loss: 0.0031 - val_acc: 0.7150
Epoch 34/100
1712/1712 [==============================] - 0s - loss: 0.0030 - acc: 0.7033 - val_loss: 0.0031 - val_acc: 0.7243
Epoch 35/100
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7039 - val_loss: 0.0041 - val_acc: 0.7243
Epoch 36/100
1712/1712 [==============================] - 0s - loss: 0.0030 - acc: 0.7079 - val_loss: 0.0034 - val_acc: 0.7150
Epoch 37/100
1712/1712 [==============================] - 0s - loss: 0.0030 - acc: 0.7109 - val_loss: 0.0030 - val_acc: 0.7243
Epoch 38/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7126 - val_loss: 0.0029 - val_acc: 0.7126
Epoch 39/100
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7091 - val_loss: 0.0029 - val_acc: 0.7243
Epoch 40/100
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7091 - val_loss: 0.0025 - val_acc: 0.7220
Epoch 41/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7126 - val_loss: 0.0029 - val_acc: 0.7453
Epoch 42/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7120 - val_loss: 0.0032 - val_acc: 0.7266
Epoch 43/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7097 - val_loss: 0.0028 - val_acc: 0.7126
Epoch 44/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7120 - val_loss: 0.0029 - val_acc: 0.7220
Epoch 45/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7091 - val_loss: 0.0026 - val_acc: 0.7383
Epoch 46/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7161 - val_loss: 0.0029 - val_acc: 0.7336
Epoch 47/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7126 - val_loss: 0.0030 - val_acc: 0.7383
Epoch 48/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7167 - val_loss: 0.0032 - val_acc: 0.7126
Epoch 49/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7220 - val_loss: 0.0032 - val_acc: 0.7547
Epoch 50/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7103 - val_loss: 0.0028 - val_acc: 0.7453
Epoch 51/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7167 - val_loss: 0.0030 - val_acc: 0.7220
Epoch 52/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7161 - val_loss: 0.0027 - val_acc: 0.7547
Epoch 53/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7126 - val_loss: 0.0026 - val_acc: 0.7313
Epoch 54/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7243 - val_loss: 0.0031 - val_acc: 0.7243
Epoch 55/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7179 - val_loss: 0.0035 - val_acc: 0.7407
Epoch 56/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7179 - val_loss: 0.0028 - val_acc: 0.7336
Epoch 57/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7144 - val_loss: 0.0036 - val_acc: 0.7126
Epoch 58/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7144 - val_loss: 0.0026 - val_acc: 0.7664
Epoch 59/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7255 - val_loss: 0.0024 - val_acc: 0.7547
Epoch 60/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7144 - val_loss: 0.0029 - val_acc: 0.7313
Epoch 61/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7243 - val_loss: 0.0028 - val_acc: 0.7103
Epoch 62/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7237 - val_loss: 0.0023 - val_acc: 0.7617
Epoch 63/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7155 - val_loss: 0.0027 - val_acc: 0.7453
Epoch 64/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7237 - val_loss: 0.0025 - val_acc: 0.7243
Epoch 65/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7220 - val_loss: 0.0028 - val_acc: 0.7290
Epoch 66/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7301 - val_loss: 0.0026 - val_acc: 0.7336
Epoch 67/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7284 - val_loss: 0.0024 - val_acc: 0.7687
Epoch 68/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7237 - val_loss: 0.0023 - val_acc: 0.7593
Epoch 69/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7278 - val_loss: 0.0025 - val_acc: 0.7383
Epoch 70/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7214 - val_loss: 0.0027 - val_acc: 0.7126
Epoch 71/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7220 - val_loss: 0.0023 - val_acc: 0.7593
Epoch 72/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7290 - val_loss: 0.0026 - val_acc: 0.7336
Epoch 73/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7307 - val_loss: 0.0026 - val_acc: 0.7477
Epoch 74/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7196 - val_loss: 0.0024 - val_acc: 0.7500
Epoch 75/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7325 - val_loss: 0.0029 - val_acc: 0.7266
Epoch 76/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7313 - val_loss: 0.0028 - val_acc: 0.7266
Epoch 77/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7243 - val_loss: 0.0021 - val_acc: 0.7593
Epoch 78/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7301 - val_loss: 0.0022 - val_acc: 0.7477
Epoch 79/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7342 - val_loss: 0.0021 - val_acc: 0.7336
Epoch 80/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7377 - val_loss: 0.0025 - val_acc: 0.7593
Epoch 81/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7278 - val_loss: 0.0022 - val_acc: 0.7593
Epoch 82/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7255 - val_loss: 0.0022 - val_acc: 0.7547
Epoch 83/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7301 - val_loss: 0.0021 - val_acc: 0.7523
Epoch 84/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7354 - val_loss: 0.0023 - val_acc: 0.7593
Epoch 85/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7331 - val_loss: 0.0024 - val_acc: 0.7547
Epoch 86/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7342 - val_loss: 0.0021 - val_acc: 0.7664
Epoch 87/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7331 - val_loss: 0.0021 - val_acc: 0.7757
Epoch 88/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7342 - val_loss: 0.0021 - val_acc: 0.7617
Epoch 89/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7348 - val_loss: 0.0029 - val_acc: 0.7640
Epoch 90/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7313 - val_loss: 0.0022 - val_acc: 0.7687
Epoch 91/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7296 - val_loss: 0.0021 - val_acc: 0.7617
Epoch 92/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7325 - val_loss: 0.0024 - val_acc: 0.7477
Epoch 93/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7237 - val_loss: 0.0020 - val_acc: 0.7664
Epoch 94/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7354 - val_loss: 0.0021 - val_acc: 0.7710
Epoch 95/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7360 - val_loss: 0.0022 - val_acc: 0.7687
Epoch 96/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7301 - val_loss: 0.0020 - val_acc: 0.7710
Epoch 97/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7331 - val_loss: 0.0023 - val_acc: 0.7734
Epoch 98/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7319 - val_loss: 0.0020 - val_acc: 0.7617
Epoch 99/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7354 - val_loss: 0.0028 - val_acc: 0.7617
Epoch 100/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7342 - val_loss: 0.0023 - val_acc: 0.7710

Adagrad¶

In [19]:

#Adagrad
model = Sequential()

model.add(Convolution2D(8, kernel_size=5, padding='same',input_shape=X_train.shape[1:]))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Convolution2D(16, kernel_size=3, padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Flatten())
model.add(Dense(30))
# Summarize the model
model.summary()

model.compile( optimizer='Adagrad', loss='mean_squared_error', metrics=['accuracy'])

checkpointer=ModelCheckpoint(save_best_only=True, filepath='adagrad_model.h5')
## TODO: Train the model
adagrad_hist = model.fit(X_train, y_train, batch_size=128, epochs=100, verbose=1, validation_split=0.2, callbacks=[checkpointer])

## TODO: Save the model as model.h5
model.save('adagrad_model.h5')

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_5 (Conv2D)            (None, 96, 96, 8)         208       
_________________________________________________________________
activation_5 (Activation)    (None, 96, 96, 8)         0         
_________________________________________________________________
max_pooling2d_5 (MaxPooling2 (None, 24, 24, 8)         0         
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 24, 24, 16)        1168      
_________________________________________________________________
activation_6 (Activation)    (None, 24, 24, 16)        0         
_________________________________________________________________
max_pooling2d_6 (MaxPooling2 (None, 6, 6, 16)          0         
_________________________________________________________________
flatten_3 (Flatten)          (None, 576)               0         
_________________________________________________________________
dense_3 (Dense)              (None, 30)                17310     
=================================================================
Total params: 18,686
Trainable params: 18,686
Non-trainable params: 0
_________________________________________________________________
Train on 1712 samples, validate on 428 samples
Epoch 1/100
1712/1712 [==============================] - 0s - loss: 0.0428 - acc: 0.4159 - val_loss: 0.0086 - val_acc: 0.6075
Epoch 2/100
1712/1712 [==============================] - 0s - loss: 0.0087 - acc: 0.6361 - val_loss: 0.0079 - val_acc: 0.5888
Epoch 3/100
1712/1712 [==============================] - 0s - loss: 0.0077 - acc: 0.6583 - val_loss: 0.0067 - val_acc: 0.6028
Epoch 4/100
1712/1712 [==============================] - 0s - loss: 0.0066 - acc: 0.6723 - val_loss: 0.0061 - val_acc: 0.6636
Epoch 5/100
1712/1712 [==============================] - 0s - loss: 0.0062 - acc: 0.6881 - val_loss: 0.0055 - val_acc: 0.6589
Epoch 6/100
1712/1712 [==============================] - 0s - loss: 0.0055 - acc: 0.6928 - val_loss: 0.0055 - val_acc: 0.6869
Epoch 7/100
1712/1712 [==============================] - 0s - loss: 0.0053 - acc: 0.6951 - val_loss: 0.0055 - val_acc: 0.6752
Epoch 8/100
1712/1712 [==============================] - 0s - loss: 0.0050 - acc: 0.6986 - val_loss: 0.0051 - val_acc: 0.6893
Epoch 9/100
1712/1712 [==============================] - 0s - loss: 0.0049 - acc: 0.6992 - val_loss: 0.0048 - val_acc: 0.6893
Epoch 10/100
1712/1712 [==============================] - 0s - loss: 0.0047 - acc: 0.6998 - val_loss: 0.0047 - val_acc: 0.6869
Epoch 11/100
1712/1712 [==============================] - 0s - loss: 0.0046 - acc: 0.7009 - val_loss: 0.0047 - val_acc: 0.6916
Epoch 12/100
1712/1712 [==============================] - 0s - loss: 0.0046 - acc: 0.6986 - val_loss: 0.0045 - val_acc: 0.6939
Epoch 13/100
1712/1712 [==============================] - 0s - loss: 0.0045 - acc: 0.6992 - val_loss: 0.0045 - val_acc: 0.6916
Epoch 14/100
1712/1712 [==============================] - 0s - loss: 0.0044 - acc: 0.7015 - val_loss: 0.0044 - val_acc: 0.6939
Epoch 15/100
1712/1712 [==============================] - 0s - loss: 0.0044 - acc: 0.6986 - val_loss: 0.0047 - val_acc: 0.6939
Epoch 16/100
1712/1712 [==============================] - 0s - loss: 0.0043 - acc: 0.7004 - val_loss: 0.0043 - val_acc: 0.6963
Epoch 17/100
1712/1712 [==============================] - 0s - loss: 0.0043 - acc: 0.7033 - val_loss: 0.0043 - val_acc: 0.6986
Epoch 18/100
1712/1712 [==============================] - 0s - loss: 0.0043 - acc: 0.7009 - val_loss: 0.0043 - val_acc: 0.6986
Epoch 19/100
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7021 - val_loss: 0.0044 - val_acc: 0.6986
Epoch 20/100
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7027 - val_loss: 0.0045 - val_acc: 0.6963
Epoch 21/100
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7009 - val_loss: 0.0042 - val_acc: 0.6963
Epoch 22/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.6986 - val_loss: 0.0042 - val_acc: 0.6963
Epoch 23/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7009 - val_loss: 0.0042 - val_acc: 0.6963
Epoch 24/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7021 - val_loss: 0.0041 - val_acc: 0.6963
Epoch 25/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7027 - val_loss: 0.0041 - val_acc: 0.6939
Epoch 26/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7027 - val_loss: 0.0042 - val_acc: 0.6986
Epoch 27/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7027 - val_loss: 0.0040 - val_acc: 0.6986
Epoch 28/100
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7004 - val_loss: 0.0040 - val_acc: 0.7009
Epoch 29/100
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.6998 - val_loss: 0.0040 - val_acc: 0.6986
Epoch 30/100
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.6998 - val_loss: 0.0040 - val_acc: 0.7009
Epoch 31/100
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7009 - val_loss: 0.0039 - val_acc: 0.7033
Epoch 32/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7009 - val_loss: 0.0040 - val_acc: 0.6963
Epoch 33/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7009 - val_loss: 0.0039 - val_acc: 0.6986
Epoch 34/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7004 - val_loss: 0.0039 - val_acc: 0.7009
Epoch 35/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7009 - val_loss: 0.0038 - val_acc: 0.7079
Epoch 36/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7004 - val_loss: 0.0038 - val_acc: 0.6986
Epoch 37/100
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.6992 - val_loss: 0.0038 - val_acc: 0.6986
Epoch 38/100
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7015 - val_loss: 0.0038 - val_acc: 0.6986
Epoch 39/100
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7021 - val_loss: 0.0038 - val_acc: 0.7009
Epoch 40/100
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7009 - val_loss: 0.0037 - val_acc: 0.7103
Epoch 41/100
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7015 - val_loss: 0.0037 - val_acc: 0.7126
Epoch 42/100
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7009 - val_loss: 0.0036 - val_acc: 0.7126
Epoch 43/100
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7009 - val_loss: 0.0036 - val_acc: 0.6963
Epoch 44/100
1712/1712 [==============================] - 0s - loss: 0.0035 - acc: 0.7033 - val_loss: 0.0036 - val_acc: 0.7126
Epoch 45/100
1712/1712 [==============================] - 0s - loss: 0.0035 - acc: 0.7021 - val_loss: 0.0036 - val_acc: 0.7056
Epoch 46/100
1712/1712 [==============================] - 0s - loss: 0.0035 - acc: 0.7056 - val_loss: 0.0036 - val_acc: 0.7126
Epoch 47/100
1712/1712 [==============================] - 0s - loss: 0.0035 - acc: 0.7027 - val_loss: 0.0036 - val_acc: 0.6939
Epoch 48/100
1712/1712 [==============================] - 0s - loss: 0.0034 - acc: 0.7039 - val_loss: 0.0035 - val_acc: 0.7173
Epoch 49/100
1712/1712 [==============================] - 0s - loss: 0.0035 - acc: 0.7062 - val_loss: 0.0035 - val_acc: 0.7103
Epoch 50/100
1712/1712 [==============================] - 0s - loss: 0.0034 - acc: 0.7056 - val_loss: 0.0034 - val_acc: 0.7173
Epoch 51/100
1712/1712 [==============================] - 0s - loss: 0.0034 - acc: 0.7039 - val_loss: 0.0034 - val_acc: 0.7196
Epoch 52/100
1712/1712 [==============================] - 0s - loss: 0.0033 - acc: 0.7015 - val_loss: 0.0034 - val_acc: 0.7103
Epoch 53/100
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7039 - val_loss: 0.0034 - val_acc: 0.7220
Epoch 54/100
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7021 - val_loss: 0.0033 - val_acc: 0.7126
Epoch 55/100
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7021 - val_loss: 0.0035 - val_acc: 0.7173
Epoch 56/100
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7015 - val_loss: 0.0032 - val_acc: 0.7173
Epoch 57/100
1712/1712 [==============================] - 0s - loss: 0.0031 - acc: 0.7074 - val_loss: 0.0032 - val_acc: 0.7173
Epoch 58/100
1712/1712 [==============================] - 0s - loss: 0.0031 - acc: 0.7050 - val_loss: 0.0034 - val_acc: 0.7150
Epoch 59/100
1712/1712 [==============================] - 0s - loss: 0.0030 - acc: 0.7062 - val_loss: 0.0031 - val_acc: 0.7103
Epoch 60/100
1712/1712 [==============================] - 0s - loss: 0.0030 - acc: 0.7097 - val_loss: 0.0030 - val_acc: 0.7126
Epoch 61/100
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7085 - val_loss: 0.0031 - val_acc: 0.7126
Epoch 62/100
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7068 - val_loss: 0.0029 - val_acc: 0.7220
Epoch 63/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7114 - val_loss: 0.0029 - val_acc: 0.7196
Epoch 64/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7114 - val_loss: 0.0029 - val_acc: 0.7173
Epoch 65/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7074 - val_loss: 0.0030 - val_acc: 0.7126
Epoch 66/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7150 - val_loss: 0.0028 - val_acc: 0.7196
Epoch 67/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7155 - val_loss: 0.0029 - val_acc: 0.7196
Epoch 68/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7144 - val_loss: 0.0028 - val_acc: 0.7150
Epoch 69/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7150 - val_loss: 0.0028 - val_acc: 0.7290
Epoch 70/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7185 - val_loss: 0.0028 - val_acc: 0.7243
Epoch 71/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7150 - val_loss: 0.0028 - val_acc: 0.7220
Epoch 72/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7144 - val_loss: 0.0031 - val_acc: 0.7196
Epoch 73/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7190 - val_loss: 0.0027 - val_acc: 0.7150
Epoch 74/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7150 - val_loss: 0.0028 - val_acc: 0.7196
Epoch 75/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7167 - val_loss: 0.0027 - val_acc: 0.7173
Epoch 76/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7208 - val_loss: 0.0028 - val_acc: 0.7290
Epoch 77/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7190 - val_loss: 0.0027 - val_acc: 0.7196
Epoch 78/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7196 - val_loss: 0.0027 - val_acc: 0.7173
Epoch 79/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7225 - val_loss: 0.0026 - val_acc: 0.7173
Epoch 80/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7243 - val_loss: 0.0026 - val_acc: 0.7336
Epoch 81/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7202 - val_loss: 0.0026 - val_acc: 0.7266
Epoch 82/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7231 - val_loss: 0.0026 - val_acc: 0.7243
Epoch 83/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7220 - val_loss: 0.0027 - val_acc: 0.7220
Epoch 84/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7243 - val_loss: 0.0026 - val_acc: 0.7220
Epoch 85/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7208 - val_loss: 0.0026 - val_acc: 0.7290
Epoch 86/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7290 - val_loss: 0.0027 - val_acc: 0.7266
Epoch 87/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7185 - val_loss: 0.0026 - val_acc: 0.7266
Epoch 88/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7278 - val_loss: 0.0025 - val_acc: 0.7266
Epoch 89/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7266 - val_loss: 0.0025 - val_acc: 0.7313
Epoch 90/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7296 - val_loss: 0.0025 - val_acc: 0.7360
Epoch 91/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7266 - val_loss: 0.0026 - val_acc: 0.7360
Epoch 92/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7255 - val_loss: 0.0025 - val_acc: 0.7290
Epoch 93/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7255 - val_loss: 0.0025 - val_acc: 0.7266
Epoch 94/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7261 - val_loss: 0.0025 - val_acc: 0.7313
Epoch 95/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7278 - val_loss: 0.0025 - val_acc: 0.7336
Epoch 96/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7290 - val_loss: 0.0025 - val_acc: 0.7266
Epoch 97/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7284 - val_loss: 0.0025 - val_acc: 0.7266
Epoch 98/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7290 - val_loss: 0.0025 - val_acc: 0.7360
Epoch 99/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7319 - val_loss: 0.0025 - val_acc: 0.7313
Epoch 100/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7272 - val_loss: 0.0025 - val_acc: 0.7360

Adadelta¶

In [20]:

#Adadelta

model = Sequential()

model.add(Convolution2D(8, kernel_size=5, padding='same',input_shape=X_train.shape[1:]))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Convolution2D(16, kernel_size=3, padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Flatten())
model.add(Dense(30))
# Summarize the model
model.summary()

model.compile( optimizer='Adadelta', loss='mean_squared_error', metrics=['accuracy'])

checkpointer=ModelCheckpoint(save_best_only=True, filepath='adadelta_model.h5')
## TODO: Train the model
adadelta_hist = model.fit(X_train, y_train, batch_size=128, epochs=100, verbose=1, validation_split=0.2, callbacks=[checkpointer])

## TODO: Save the model as model.h5
model.save('adadelta_model.h5')

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_7 (Conv2D)            (None, 96, 96, 8)         208       
_________________________________________________________________
activation_7 (Activation)    (None, 96, 96, 8)         0         
_________________________________________________________________
max_pooling2d_7 (MaxPooling2 (None, 24, 24, 8)         0         
_________________________________________________________________
conv2d_8 (Conv2D)            (None, 24, 24, 16)        1168      
_________________________________________________________________
activation_8 (Activation)    (None, 24, 24, 16)        0         
_________________________________________________________________
max_pooling2d_8 (MaxPooling2 (None, 6, 6, 16)          0         
_________________________________________________________________
flatten_4 (Flatten)          (None, 576)               0         
_________________________________________________________________
dense_4 (Dense)              (None, 30)                17310     
=================================================================
Total params: 18,686
Trainable params: 18,686
Non-trainable params: 0
_________________________________________________________________
Train on 1712 samples, validate on 428 samples
Epoch 1/100
1712/1712 [==============================] - 0s - loss: 0.1097 - acc: 0.6565 - val_loss: 0.0420 - val_acc: 0.6799
Epoch 2/100
1712/1712 [==============================] - 0s - loss: 0.0190 - acc: 0.6355 - val_loss: 0.0131 - val_acc: 0.6285
Epoch 3/100
1712/1712 [==============================] - 0s - loss: 0.0118 - acc: 0.5905 - val_loss: 0.0109 - val_acc: 0.6355
Epoch 4/100
1712/1712 [==============================] - 0s - loss: 0.0108 - acc: 0.5940 - val_loss: 0.0099 - val_acc: 0.6308
Epoch 5/100
1712/1712 [==============================] - 0s - loss: 0.0100 - acc: 0.5975 - val_loss: 0.0092 - val_acc: 0.6192
Epoch 6/100
1712/1712 [==============================] - 0s - loss: 0.0093 - acc: 0.5987 - val_loss: 0.0085 - val_acc: 0.6168
Epoch 7/100
1712/1712 [==============================] - 0s - loss: 0.0085 - acc: 0.6040 - val_loss: 0.0089 - val_acc: 0.6215
Epoch 8/100
1712/1712 [==============================] - 0s - loss: 0.0080 - acc: 0.6192 - val_loss: 0.0075 - val_acc: 0.6495
Epoch 9/100
1712/1712 [==============================] - 0s - loss: 0.0075 - acc: 0.6332 - val_loss: 0.0077 - val_acc: 0.6379
Epoch 10/100
1712/1712 [==============================] - 0s - loss: 0.0072 - acc: 0.6379 - val_loss: 0.0070 - val_acc: 0.6682
Epoch 11/100
1712/1712 [==============================] - 0s - loss: 0.0068 - acc: 0.6525 - val_loss: 0.0067 - val_acc: 0.6706
Epoch 12/100
1712/1712 [==============================] - 0s - loss: 0.0064 - acc: 0.6612 - val_loss: 0.0061 - val_acc: 0.6285
Epoch 13/100
1712/1712 [==============================] - 0s - loss: 0.0061 - acc: 0.6671 - val_loss: 0.0059 - val_acc: 0.6916
Epoch 14/100
1712/1712 [==============================] - 0s - loss: 0.0059 - acc: 0.6805 - val_loss: 0.0059 - val_acc: 0.6963
Epoch 15/100
1712/1712 [==============================] - 0s - loss: 0.0063 - acc: 0.6928 - val_loss: 0.0071 - val_acc: 0.6986
Epoch 16/100
1712/1712 [==============================] - 0s - loss: 0.0058 - acc: 0.6933 - val_loss: 0.0055 - val_acc: 0.6963
Epoch 17/100
1712/1712 [==============================] - 0s - loss: 0.0056 - acc: 0.6986 - val_loss: 0.0059 - val_acc: 0.6963
Epoch 18/100
1712/1712 [==============================] - 0s - loss: 0.0055 - acc: 0.7027 - val_loss: 0.0056 - val_acc: 0.6939
Epoch 19/100
1712/1712 [==============================] - 0s - loss: 0.0056 - acc: 0.7056 - val_loss: 0.0060 - val_acc: 0.6986
Epoch 20/100
1712/1712 [==============================] - 0s - loss: 0.0055 - acc: 0.7015 - val_loss: 0.0053 - val_acc: 0.6986
Epoch 21/100
1712/1712 [==============================] - 0s - loss: 0.0054 - acc: 0.7068 - val_loss: 0.0052 - val_acc: 0.7033
Epoch 22/100
1712/1712 [==============================] - 0s - loss: 0.0054 - acc: 0.7056 - val_loss: 0.0051 - val_acc: 0.6986
Epoch 23/100
1712/1712 [==============================] - 0s - loss: 0.0052 - acc: 0.7068 - val_loss: 0.0050 - val_acc: 0.7009
Epoch 24/100
1712/1712 [==============================] - 0s - loss: 0.0052 - acc: 0.7062 - val_loss: 0.0055 - val_acc: 0.7033
Epoch 25/100
1712/1712 [==============================] - 0s - loss: 0.0051 - acc: 0.7056 - val_loss: 0.0058 - val_acc: 0.6916
Epoch 26/100
1712/1712 [==============================] - 0s - loss: 0.0051 - acc: 0.7085 - val_loss: 0.0062 - val_acc: 0.7033
Epoch 27/100
1712/1712 [==============================] - 0s - loss: 0.0052 - acc: 0.7085 - val_loss: 0.0048 - val_acc: 0.7009
Epoch 28/100
1712/1712 [==============================] - 0s - loss: 0.0049 - acc: 0.7079 - val_loss: 0.0049 - val_acc: 0.7033
Epoch 29/100
1712/1712 [==============================] - 0s - loss: 0.0050 - acc: 0.7062 - val_loss: 0.0058 - val_acc: 0.7009
Epoch 30/100
1712/1712 [==============================] - 0s - loss: 0.0049 - acc: 0.7079 - val_loss: 0.0053 - val_acc: 0.6916
Epoch 31/100
1712/1712 [==============================] - 0s - loss: 0.0049 - acc: 0.7074 - val_loss: 0.0052 - val_acc: 0.7056
Epoch 32/100
1712/1712 [==============================] - 0s - loss: 0.0049 - acc: 0.7056 - val_loss: 0.0048 - val_acc: 0.7009
Epoch 33/100
1712/1712 [==============================] - 0s - loss: 0.0050 - acc: 0.7074 - val_loss: 0.0047 - val_acc: 0.7009
Epoch 34/100
1712/1712 [==============================] - 0s - loss: 0.0048 - acc: 0.7068 - val_loss: 0.0046 - val_acc: 0.7009
Epoch 35/100
1712/1712 [==============================] - 0s - loss: 0.0047 - acc: 0.7068 - val_loss: 0.0050 - val_acc: 0.6986
Epoch 36/100
1712/1712 [==============================] - 0s - loss: 0.0048 - acc: 0.7091 - val_loss: 0.0045 - val_acc: 0.7033
Epoch 37/100
1712/1712 [==============================] - 0s - loss: 0.0045 - acc: 0.7085 - val_loss: 0.0047 - val_acc: 0.6986
Epoch 38/100
1712/1712 [==============================] - 0s - loss: 0.0046 - acc: 0.7091 - val_loss: 0.0047 - val_acc: 0.7033
Epoch 39/100
1712/1712 [==============================] - 0s - loss: 0.0046 - acc: 0.7079 - val_loss: 0.0045 - val_acc: 0.7033
Epoch 40/100
1712/1712 [==============================] - 0s - loss: 0.0045 - acc: 0.7074 - val_loss: 0.0044 - val_acc: 0.7033
Epoch 41/100
1712/1712 [==============================] - 0s - loss: 0.0048 - acc: 0.7079 - val_loss: 0.0049 - val_acc: 0.6986
Epoch 42/100
1712/1712 [==============================] - 0s - loss: 0.0047 - acc: 0.7091 - val_loss: 0.0045 - val_acc: 0.7033
Epoch 43/100
1712/1712 [==============================] - 0s - loss: 0.0046 - acc: 0.7079 - val_loss: 0.0051 - val_acc: 0.7033
Epoch 44/100
1712/1712 [==============================] - 0s - loss: 0.0046 - acc: 0.7079 - val_loss: 0.0047 - val_acc: 0.7033
Epoch 45/100
1712/1712 [==============================] - 0s - loss: 0.0043 - acc: 0.7091 - val_loss: 0.0044 - val_acc: 0.7033
Epoch 46/100
1712/1712 [==============================] - 0s - loss: 0.0044 - acc: 0.7091 - val_loss: 0.0043 - val_acc: 0.7033
Epoch 47/100
1712/1712 [==============================] - 0s - loss: 0.0044 - acc: 0.7079 - val_loss: 0.0045 - val_acc: 0.7033
Epoch 48/100
1712/1712 [==============================] - 0s - loss: 0.0045 - acc: 0.7085 - val_loss: 0.0045 - val_acc: 0.7033
Epoch 49/100
1712/1712 [==============================] - 0s - loss: 0.0043 - acc: 0.7085 - val_loss: 0.0048 - val_acc: 0.6963
Epoch 50/100
1712/1712 [==============================] - 0s - loss: 0.0044 - acc: 0.7074 - val_loss: 0.0043 - val_acc: 0.7033
Epoch 51/100
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7085 - val_loss: 0.0042 - val_acc: 0.6986
Epoch 52/100
1712/1712 [==============================] - 0s - loss: 0.0043 - acc: 0.7097 - val_loss: 0.0044 - val_acc: 0.7033
Epoch 53/100
1712/1712 [==============================] - 0s - loss: 0.0044 - acc: 0.7079 - val_loss: 0.0042 - val_acc: 0.7009
Epoch 54/100
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7085 - val_loss: 0.0045 - val_acc: 0.6986
Epoch 55/100
1712/1712 [==============================] - 0s - loss: 0.0043 - acc: 0.7079 - val_loss: 0.0043 - val_acc: 0.7009
Epoch 56/100
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7097 - val_loss: 0.0044 - val_acc: 0.7009
Epoch 57/100
1712/1712 [==============================] - 0s - loss: 0.0044 - acc: 0.7097 - val_loss: 0.0042 - val_acc: 0.7033
Epoch 58/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7097 - val_loss: 0.0043 - val_acc: 0.6986
Epoch 59/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7097 - val_loss: 0.0043 - val_acc: 0.7009
Epoch 60/100
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7085 - val_loss: 0.0042 - val_acc: 0.7009
Epoch 61/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7091 - val_loss: 0.0041 - val_acc: 0.7033
Epoch 62/100
1712/1712 [==============================] - 0s - loss: 0.0043 - acc: 0.7062 - val_loss: 0.0042 - val_acc: 0.6986
Epoch 63/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7109 - val_loss: 0.0041 - val_acc: 0.7056
Epoch 64/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7114 - val_loss: 0.0042 - val_acc: 0.7009
Epoch 65/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7097 - val_loss: 0.0042 - val_acc: 0.7009
Epoch 66/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7079 - val_loss: 0.0041 - val_acc: 0.7033
Epoch 67/100
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7109 - val_loss: 0.0041 - val_acc: 0.7033
Epoch 68/100
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7103 - val_loss: 0.0044 - val_acc: 0.7009
Epoch 69/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7109 - val_loss: 0.0044 - val_acc: 0.7009
Epoch 70/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7097 - val_loss: 0.0044 - val_acc: 0.7009
Epoch 71/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7120 - val_loss: 0.0041 - val_acc: 0.7056
Epoch 72/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7109 - val_loss: 0.0040 - val_acc: 0.7033
Epoch 73/100
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7109 - val_loss: 0.0039 - val_acc: 0.7033
Epoch 74/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7103 - val_loss: 0.0040 - val_acc: 0.7009
Epoch 75/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7120 - val_loss: 0.0040 - val_acc: 0.6986
Epoch 76/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7114 - val_loss: 0.0042 - val_acc: 0.7103
Epoch 77/100
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7114 - val_loss: 0.0039 - val_acc: 0.7009
Epoch 78/100
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7132 - val_loss: 0.0043 - val_acc: 0.7033
Epoch 79/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7126 - val_loss: 0.0040 - val_acc: 0.7033
Epoch 80/100
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7126 - val_loss: 0.0038 - val_acc: 0.7009
Epoch 81/100
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7126 - val_loss: 0.0039 - val_acc: 0.7009
Epoch 82/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7114 - val_loss: 0.0038 - val_acc: 0.7056
Epoch 83/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7103 - val_loss: 0.0039 - val_acc: 0.7033
Epoch 84/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7114 - val_loss: 0.0044 - val_acc: 0.7009
Epoch 85/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7144 - val_loss: 0.0038 - val_acc: 0.7056
Epoch 86/100
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7091 - val_loss: 0.0038 - val_acc: 0.7079
Epoch 87/100
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7103 - val_loss: 0.0040 - val_acc: 0.7079
Epoch 88/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7103 - val_loss: 0.0038 - val_acc: 0.7079
Epoch 89/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7132 - val_loss: 0.0040 - val_acc: 0.7056
Epoch 90/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7150 - val_loss: 0.0040 - val_acc: 0.7056
Epoch 91/100
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7085 - val_loss: 0.0037 - val_acc: 0.7079
Epoch 92/100
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7114 - val_loss: 0.0037 - val_acc: 0.7079
Epoch 93/100
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7126 - val_loss: 0.0038 - val_acc: 0.7079
Epoch 94/100
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7091 - val_loss: 0.0039 - val_acc: 0.7126
Epoch 95/100
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7114 - val_loss: 0.0037 - val_acc: 0.7056
Epoch 96/100
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7114 - val_loss: 0.0040 - val_acc: 0.7056
Epoch 97/100
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7126 - val_loss: 0.0038 - val_acc: 0.7033
Epoch 98/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7120 - val_loss: 0.0041 - val_acc: 0.7056
Epoch 99/100
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7132 - val_loss: 0.0038 - val_acc: 0.7056
Epoch 100/100
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7161 - val_loss: 0.0037 - val_acc: 0.7009

Adam¶

In [21]:

#Adam

model = Sequential()

model.add(Convolution2D(8, kernel_size=5, padding='same',input_shape=X_train.shape[1:]))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Convolution2D(16, kernel_size=3, padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Flatten())
model.add(Dense(30))
# Summarize the model
model.summary()

model.compile( optimizer='Adam', loss='mean_squared_error', metrics=['accuracy'])

checkpointer=ModelCheckpoint(save_best_only=True, filepath='adam_model.h5')
## TODO: Train the model
adam_hist = model.fit(X_train, y_train, batch_size=128, epochs=100, verbose=1, validation_split=0.2, callbacks=[checkpointer])

## TODO: Save the model as model.h5
model.save('adam_model.h5')

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_9 (Conv2D)            (None, 96, 96, 8)         208       
_________________________________________________________________
activation_9 (Activation)    (None, 96, 96, 8)         0         
_________________________________________________________________
max_pooling2d_9 (MaxPooling2 (None, 24, 24, 8)         0         
_________________________________________________________________
conv2d_10 (Conv2D)           (None, 24, 24, 16)        1168      
_________________________________________________________________
activation_10 (Activation)   (None, 24, 24, 16)        0         
_________________________________________________________________
max_pooling2d_10 (MaxPooling (None, 6, 6, 16)          0         
_________________________________________________________________
flatten_5 (Flatten)          (None, 576)               0         
_________________________________________________________________
dense_5 (Dense)              (None, 30)                17310     
=================================================================
Total params: 18,686
Trainable params: 18,686
Non-trainable params: 0
_________________________________________________________________
Train on 1712 samples, validate on 428 samples
Epoch 1/100
1712/1712 [==============================] - 0s - loss: 0.0638 - acc: 0.6735 - val_loss: 0.0173 - val_acc: 0.6916
Epoch 2/100
1712/1712 [==============================] - 0s - loss: 0.0151 - acc: 0.4083 - val_loss: 0.0110 - val_acc: 0.2523
Epoch 3/100
1712/1712 [==============================] - 0s - loss: 0.0109 - acc: 0.4725 - val_loss: 0.0096 - val_acc: 0.6799
Epoch 4/100
1712/1712 [==============================] - 0s - loss: 0.0089 - acc: 0.6933 - val_loss: 0.0076 - val_acc: 0.6729
Epoch 5/100
1712/1712 [==============================] - 0s - loss: 0.0072 - acc: 0.6525 - val_loss: 0.0064 - val_acc: 0.6589
Epoch 6/100
1712/1712 [==============================] - 0s - loss: 0.0060 - acc: 0.6904 - val_loss: 0.0055 - val_acc: 0.6916
Epoch 7/100
1712/1712 [==============================] - 0s - loss: 0.0051 - acc: 0.6980 - val_loss: 0.0049 - val_acc: 0.6869
Epoch 8/100
1712/1712 [==============================] - 0s - loss: 0.0047 - acc: 0.7033 - val_loss: 0.0045 - val_acc: 0.6916
Epoch 9/100
1712/1712 [==============================] - 0s - loss: 0.0043 - acc: 0.7033 - val_loss: 0.0043 - val_acc: 0.6963
Epoch 10/100
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7068 - val_loss: 0.0042 - val_acc: 0.6939
Epoch 11/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7079 - val_loss: 0.0039 - val_acc: 0.7033
Epoch 12/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7091 - val_loss: 0.0038 - val_acc: 0.7103
Epoch 13/100
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7097 - val_loss: 0.0036 - val_acc: 0.7079
Epoch 14/100
1712/1712 [==============================] - 0s - loss: 0.0034 - acc: 0.7085 - val_loss: 0.0034 - val_acc: 0.7126
Epoch 15/100
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7097 - val_loss: 0.0033 - val_acc: 0.6939
Epoch 16/100
1712/1712 [==============================] - 0s - loss: 0.0031 - acc: 0.7062 - val_loss: 0.0031 - val_acc: 0.7056
Epoch 17/100
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7085 - val_loss: 0.0030 - val_acc: 0.7103
Epoch 18/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7138 - val_loss: 0.0029 - val_acc: 0.7126
Epoch 19/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7097 - val_loss: 0.0028 - val_acc: 0.7196
Epoch 20/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7114 - val_loss: 0.0027 - val_acc: 0.7150
Epoch 21/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7120 - val_loss: 0.0028 - val_acc: 0.7173
Epoch 22/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7097 - val_loss: 0.0026 - val_acc: 0.7220
Epoch 23/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7079 - val_loss: 0.0026 - val_acc: 0.7266
Epoch 24/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7144 - val_loss: 0.0027 - val_acc: 0.7220
Epoch 25/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7120 - val_loss: 0.0025 - val_acc: 0.7266
Epoch 26/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7167 - val_loss: 0.0026 - val_acc: 0.7313
Epoch 27/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7185 - val_loss: 0.0025 - val_acc: 0.7266
Epoch 28/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7167 - val_loss: 0.0025 - val_acc: 0.7266
Epoch 29/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7220 - val_loss: 0.0025 - val_acc: 0.7173
Epoch 30/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7167 - val_loss: 0.0024 - val_acc: 0.7196
Epoch 31/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7249 - val_loss: 0.0024 - val_acc: 0.7196
Epoch 32/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7202 - val_loss: 0.0024 - val_acc: 0.7313
Epoch 33/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7190 - val_loss: 0.0024 - val_acc: 0.7336
Epoch 34/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7208 - val_loss: 0.0024 - val_acc: 0.7360
Epoch 35/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7225 - val_loss: 0.0023 - val_acc: 0.7266
Epoch 36/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7185 - val_loss: 0.0023 - val_acc: 0.7220
Epoch 37/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7237 - val_loss: 0.0023 - val_acc: 0.7290
Epoch 38/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7266 - val_loss: 0.0023 - val_acc: 0.7430
Epoch 39/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7272 - val_loss: 0.0023 - val_acc: 0.7430
Epoch 40/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7220 - val_loss: 0.0023 - val_acc: 0.7196
Epoch 41/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7336 - val_loss: 0.0023 - val_acc: 0.7220
Epoch 42/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7249 - val_loss: 0.0022 - val_acc: 0.7313
Epoch 43/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7249 - val_loss: 0.0023 - val_acc: 0.7360
Epoch 44/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7284 - val_loss: 0.0023 - val_acc: 0.7383
Epoch 45/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7249 - val_loss: 0.0022 - val_acc: 0.7407
Epoch 46/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7214 - val_loss: 0.0023 - val_acc: 0.7360
Epoch 47/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7296 - val_loss: 0.0022 - val_acc: 0.7407
Epoch 48/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7272 - val_loss: 0.0022 - val_acc: 0.7407
Epoch 49/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7261 - val_loss: 0.0023 - val_acc: 0.7336
Epoch 50/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7237 - val_loss: 0.0022 - val_acc: 0.7290
Epoch 51/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7278 - val_loss: 0.0022 - val_acc: 0.7360
Epoch 52/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7296 - val_loss: 0.0022 - val_acc: 0.7383
Epoch 53/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7348 - val_loss: 0.0021 - val_acc: 0.7360
Epoch 54/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7301 - val_loss: 0.0021 - val_acc: 0.7313
Epoch 55/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7313 - val_loss: 0.0022 - val_acc: 0.7360
Epoch 56/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7278 - val_loss: 0.0021 - val_acc: 0.7407
Epoch 57/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7261 - val_loss: 0.0022 - val_acc: 0.7430
Epoch 58/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7319 - val_loss: 0.0022 - val_acc: 0.7407
Epoch 59/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7301 - val_loss: 0.0021 - val_acc: 0.7407
Epoch 60/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7313 - val_loss: 0.0021 - val_acc: 0.7383
Epoch 61/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7331 - val_loss: 0.0021 - val_acc: 0.7430
Epoch 62/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7331 - val_loss: 0.0021 - val_acc: 0.7477
Epoch 63/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7290 - val_loss: 0.0021 - val_acc: 0.7523
Epoch 64/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7307 - val_loss: 0.0021 - val_acc: 0.7477
Epoch 65/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7325 - val_loss: 0.0021 - val_acc: 0.7453
Epoch 66/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7301 - val_loss: 0.0021 - val_acc: 0.7453
Epoch 67/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7301 - val_loss: 0.0021 - val_acc: 0.7453
Epoch 68/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7313 - val_loss: 0.0021 - val_acc: 0.7453
Epoch 69/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7301 - val_loss: 0.0021 - val_acc: 0.7547
Epoch 70/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7325 - val_loss: 0.0020 - val_acc: 0.7500
Epoch 71/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7331 - val_loss: 0.0020 - val_acc: 0.7523
Epoch 72/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7325 - val_loss: 0.0020 - val_acc: 0.7547
Epoch 73/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7296 - val_loss: 0.0021 - val_acc: 0.7500
Epoch 74/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7407 - val_loss: 0.0021 - val_acc: 0.7500
Epoch 75/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7366 - val_loss: 0.0020 - val_acc: 0.7570
Epoch 76/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7360 - val_loss: 0.0020 - val_acc: 0.7477
Epoch 77/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7331 - val_loss: 0.0020 - val_acc: 0.7500
Epoch 78/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7366 - val_loss: 0.0020 - val_acc: 0.7570
Epoch 79/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7482 - val_loss: 0.0020 - val_acc: 0.7523
Epoch 80/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7447 - val_loss: 0.0020 - val_acc: 0.7477
Epoch 81/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7401 - val_loss: 0.0020 - val_acc: 0.7523
Epoch 82/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7377 - val_loss: 0.0020 - val_acc: 0.7570
Epoch 83/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7401 - val_loss: 0.0020 - val_acc: 0.7523
Epoch 84/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7418 - val_loss: 0.0020 - val_acc: 0.7617
Epoch 85/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7418 - val_loss: 0.0020 - val_acc: 0.7570
Epoch 86/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7412 - val_loss: 0.0020 - val_acc: 0.7570
Epoch 87/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7395 - val_loss: 0.0020 - val_acc: 0.7617
Epoch 88/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7383 - val_loss: 0.0020 - val_acc: 0.7547
Epoch 89/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7442 - val_loss: 0.0020 - val_acc: 0.7617
Epoch 90/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7389 - val_loss: 0.0020 - val_acc: 0.7664
Epoch 91/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7418 - val_loss: 0.0020 - val_acc: 0.7664
Epoch 92/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7453 - val_loss: 0.0021 - val_acc: 0.7710
Epoch 93/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7412 - val_loss: 0.0020 - val_acc: 0.7617
Epoch 94/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7447 - val_loss: 0.0020 - val_acc: 0.7710
Epoch 95/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7442 - val_loss: 0.0019 - val_acc: 0.7664
Epoch 96/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7424 - val_loss: 0.0019 - val_acc: 0.7734
Epoch 97/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7477 - val_loss: 0.0020 - val_acc: 0.7664
Epoch 98/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7436 - val_loss: 0.0019 - val_acc: 0.7664
Epoch 99/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7395 - val_loss: 0.0019 - val_acc: 0.7757
Epoch 100/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7471 - val_loss: 0.0019 - val_acc: 0.7780

Adamax¶

In [22]:

#Adamax

model = Sequential()

model.add(Convolution2D(8, kernel_size=5, padding='same',input_shape=X_train.shape[1:]))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Convolution2D(16, kernel_size=3, padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Flatten())
model.add(Dense(30))
# Summarize the model
model.summary()

model.compile( optimizer='Adamax', loss='mean_squared_error', metrics=['accuracy'])

checkpointer=ModelCheckpoint(save_best_only=True, filepath='adamax_model.h5')
## TODO: Train the model
adamax_hist = model.fit(X_train, y_train, batch_size=128, epochs=100, verbose=1, validation_split=0.2, callbacks=[checkpointer])

## TODO: Save the model as model.h5
model.save('adamax_model.h5')

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_11 (Conv2D)           (None, 96, 96, 8)         208       
_________________________________________________________________
activation_11 (Activation)   (None, 96, 96, 8)         0         
_________________________________________________________________
max_pooling2d_11 (MaxPooling (None, 24, 24, 8)         0         
_________________________________________________________________
conv2d_12 (Conv2D)           (None, 24, 24, 16)        1168      
_________________________________________________________________
activation_12 (Activation)   (None, 24, 24, 16)        0         
_________________________________________________________________
max_pooling2d_12 (MaxPooling (None, 6, 6, 16)          0         
_________________________________________________________________
flatten_6 (Flatten)          (None, 576)               0         
_________________________________________________________________
dense_6 (Dense)              (None, 30)                17310     
=================================================================
Total params: 18,686
Trainable params: 18,686
Non-trainable params: 0
_________________________________________________________________
Train on 1712 samples, validate on 428 samples
Epoch 1/100
1712/1712 [==============================] - 0s - loss: 0.0528 - acc: 0.1016 - val_loss: 0.0151 - val_acc: 0.6776
Epoch 2/100
1712/1712 [==============================] - 0s - loss: 0.0141 - acc: 0.7056 - val_loss: 0.0107 - val_acc: 0.6963
Epoch 3/100
1712/1712 [==============================] - 0s - loss: 0.0106 - acc: 0.7062 - val_loss: 0.0092 - val_acc: 0.6963
Epoch 4/100
1712/1712 [==============================] - 0s - loss: 0.0093 - acc: 0.6600 - val_loss: 0.0083 - val_acc: 0.6659
Epoch 5/100
1712/1712 [==============================] - 0s - loss: 0.0085 - acc: 0.6682 - val_loss: 0.0077 - val_acc: 0.6939
Epoch 6/100
1712/1712 [==============================] - 0s - loss: 0.0078 - acc: 0.6957 - val_loss: 0.0071 - val_acc: 0.6986
Epoch 7/100
1712/1712 [==============================] - 0s - loss: 0.0071 - acc: 0.6986 - val_loss: 0.0065 - val_acc: 0.6963
Epoch 8/100
1712/1712 [==============================] - 0s - loss: 0.0065 - acc: 0.6986 - val_loss: 0.0062 - val_acc: 0.7009
Epoch 9/100
1712/1712 [==============================] - 0s - loss: 0.0060 - acc: 0.7004 - val_loss: 0.0058 - val_acc: 0.7033
Epoch 10/100
1712/1712 [==============================] - 0s - loss: 0.0056 - acc: 0.7033 - val_loss: 0.0055 - val_acc: 0.6986
Epoch 11/100
1712/1712 [==============================] - 0s - loss: 0.0053 - acc: 0.7039 - val_loss: 0.0052 - val_acc: 0.6986
Epoch 12/100
1712/1712 [==============================] - 0s - loss: 0.0050 - acc: 0.7062 - val_loss: 0.0050 - val_acc: 0.7009
Epoch 13/100
1712/1712 [==============================] - 0s - loss: 0.0048 - acc: 0.7068 - val_loss: 0.0048 - val_acc: 0.6939
Epoch 14/100
1712/1712 [==============================] - 0s - loss: 0.0046 - acc: 0.7050 - val_loss: 0.0046 - val_acc: 0.6963
Epoch 15/100
1712/1712 [==============================] - 0s - loss: 0.0044 - acc: 0.7044 - val_loss: 0.0045 - val_acc: 0.6963
Epoch 16/100
1712/1712 [==============================] - 0s - loss: 0.0043 - acc: 0.7044 - val_loss: 0.0043 - val_acc: 0.6963
Epoch 17/100
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7068 - val_loss: 0.0042 - val_acc: 0.6939
Epoch 18/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7021 - val_loss: 0.0041 - val_acc: 0.7009
Epoch 19/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7027 - val_loss: 0.0040 - val_acc: 0.6963
Epoch 20/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7085 - val_loss: 0.0039 - val_acc: 0.6846
Epoch 21/100
1712/1712 [==============================] - 0s - loss: 0.0038 - acc: 0.7056 - val_loss: 0.0038 - val_acc: 0.7009
Epoch 22/100
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7015 - val_loss: 0.0037 - val_acc: 0.6939
Epoch 23/100
1712/1712 [==============================] - 0s - loss: 0.0035 - acc: 0.7044 - val_loss: 0.0036 - val_acc: 0.6963
Epoch 24/100
1712/1712 [==============================] - 0s - loss: 0.0035 - acc: 0.7050 - val_loss: 0.0035 - val_acc: 0.6963
Epoch 25/100
1712/1712 [==============================] - 0s - loss: 0.0034 - acc: 0.7068 - val_loss: 0.0034 - val_acc: 0.7009
Epoch 26/100
1712/1712 [==============================] - 0s - loss: 0.0033 - acc: 0.7085 - val_loss: 0.0033 - val_acc: 0.7033
Epoch 27/100
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7126 - val_loss: 0.0033 - val_acc: 0.7056
Epoch 28/100
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7138 - val_loss: 0.0032 - val_acc: 0.7103
Epoch 29/100
1712/1712 [==============================] - 0s - loss: 0.0031 - acc: 0.7190 - val_loss: 0.0031 - val_acc: 0.7056
Epoch 30/100
1712/1712 [==============================] - 0s - loss: 0.0030 - acc: 0.7120 - val_loss: 0.0031 - val_acc: 0.7150
Epoch 31/100
1712/1712 [==============================] - 0s - loss: 0.0030 - acc: 0.7150 - val_loss: 0.0030 - val_acc: 0.7173
Epoch 32/100
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7161 - val_loss: 0.0030 - val_acc: 0.7220
Epoch 33/100
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7161 - val_loss: 0.0030 - val_acc: 0.7290
Epoch 34/100
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7161 - val_loss: 0.0030 - val_acc: 0.7243
Epoch 35/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7173 - val_loss: 0.0029 - val_acc: 0.7196
Epoch 36/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7155 - val_loss: 0.0029 - val_acc: 0.7290
Epoch 37/100
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7161 - val_loss: 0.0028 - val_acc: 0.7313
Epoch 38/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7208 - val_loss: 0.0028 - val_acc: 0.7196
Epoch 39/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7214 - val_loss: 0.0028 - val_acc: 0.7220
Epoch 40/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7173 - val_loss: 0.0028 - val_acc: 0.7290
Epoch 41/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7225 - val_loss: 0.0027 - val_acc: 0.7220
Epoch 42/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7249 - val_loss: 0.0027 - val_acc: 0.7360
Epoch 43/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7179 - val_loss: 0.0027 - val_acc: 0.7453
Epoch 44/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7196 - val_loss: 0.0026 - val_acc: 0.7383
Epoch 45/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7208 - val_loss: 0.0026 - val_acc: 0.7313
Epoch 46/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7208 - val_loss: 0.0026 - val_acc: 0.7336
Epoch 47/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7190 - val_loss: 0.0026 - val_acc: 0.7290
Epoch 48/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7261 - val_loss: 0.0027 - val_acc: 0.7383
Epoch 49/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7202 - val_loss: 0.0025 - val_acc: 0.7336
Epoch 50/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7255 - val_loss: 0.0025 - val_acc: 0.7407
Epoch 51/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7284 - val_loss: 0.0025 - val_acc: 0.7453
Epoch 52/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7255 - val_loss: 0.0025 - val_acc: 0.7453
Epoch 53/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7301 - val_loss: 0.0025 - val_acc: 0.7523
Epoch 54/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7249 - val_loss: 0.0025 - val_acc: 0.7477
Epoch 55/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7237 - val_loss: 0.0024 - val_acc: 0.7453
Epoch 56/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7249 - val_loss: 0.0024 - val_acc: 0.7430
Epoch 57/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7278 - val_loss: 0.0025 - val_acc: 0.7430
Epoch 58/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7231 - val_loss: 0.0024 - val_acc: 0.7500
Epoch 59/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7296 - val_loss: 0.0024 - val_acc: 0.7477
Epoch 60/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7278 - val_loss: 0.0024 - val_acc: 0.7383
Epoch 61/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7307 - val_loss: 0.0024 - val_acc: 0.7477
Epoch 62/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7296 - val_loss: 0.0023 - val_acc: 0.7477
Epoch 63/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7325 - val_loss: 0.0024 - val_acc: 0.7453
Epoch 64/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7284 - val_loss: 0.0024 - val_acc: 0.7500
Epoch 65/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7319 - val_loss: 0.0023 - val_acc: 0.7407
Epoch 66/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7325 - val_loss: 0.0023 - val_acc: 0.7453
Epoch 67/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7342 - val_loss: 0.0023 - val_acc: 0.7383
Epoch 68/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7307 - val_loss: 0.0023 - val_acc: 0.7523
Epoch 69/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7307 - val_loss: 0.0024 - val_acc: 0.7500
Epoch 70/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7331 - val_loss: 0.0023 - val_acc: 0.7477
Epoch 71/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7319 - val_loss: 0.0023 - val_acc: 0.7523
Epoch 72/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7307 - val_loss: 0.0023 - val_acc: 0.7477
Epoch 73/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7331 - val_loss: 0.0022 - val_acc: 0.7430
Epoch 74/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7336 - val_loss: 0.0022 - val_acc: 0.7430
Epoch 75/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7290 - val_loss: 0.0022 - val_acc: 0.7453
Epoch 76/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7313 - val_loss: 0.0022 - val_acc: 0.7523
Epoch 77/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7342 - val_loss: 0.0022 - val_acc: 0.7477
Epoch 78/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7348 - val_loss: 0.0022 - val_acc: 0.7453
Epoch 79/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7331 - val_loss: 0.0022 - val_acc: 0.7430
Epoch 80/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7307 - val_loss: 0.0022 - val_acc: 0.7383
Epoch 81/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7354 - val_loss: 0.0022 - val_acc: 0.7453
Epoch 82/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7336 - val_loss: 0.0022 - val_acc: 0.7453
Epoch 83/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7336 - val_loss: 0.0022 - val_acc: 0.7500
Epoch 84/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7325 - val_loss: 0.0022 - val_acc: 0.7430
Epoch 85/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7331 - val_loss: 0.0021 - val_acc: 0.7266
Epoch 86/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7278 - val_loss: 0.0021 - val_acc: 0.7290
Epoch 87/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7342 - val_loss: 0.0022 - val_acc: 0.7313
Epoch 88/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7371 - val_loss: 0.0021 - val_acc: 0.7336
Epoch 89/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7325 - val_loss: 0.0021 - val_acc: 0.7430
Epoch 90/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7331 - val_loss: 0.0021 - val_acc: 0.7336
Epoch 91/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7371 - val_loss: 0.0021 - val_acc: 0.7360
Epoch 92/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7377 - val_loss: 0.0021 - val_acc: 0.7360
Epoch 93/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7301 - val_loss: 0.0022 - val_acc: 0.7407
Epoch 94/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7401 - val_loss: 0.0021 - val_acc: 0.7336
Epoch 95/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7348 - val_loss: 0.0021 - val_acc: 0.7313
Epoch 96/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7377 - val_loss: 0.0021 - val_acc: 0.7243
Epoch 97/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7336 - val_loss: 0.0021 - val_acc: 0.7243
Epoch 98/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7319 - val_loss: 0.0021 - val_acc: 0.7313
Epoch 99/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7348 - val_loss: 0.0020 - val_acc: 0.7313
Epoch 100/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7418 - val_loss: 0.0021 - val_acc: 0.7430

Nadam¶

In [23]:

#Nadam

model = Sequential()

model.add(Convolution2D(8, kernel_size=5, padding='same',input_shape=X_train.shape[1:]))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Convolution2D(16, kernel_size=3, padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Flatten())
model.add(Dense(30))
# Summarize the model
model.summary()

model.compile( optimizer='Nadam', loss='mean_squared_error', metrics=['accuracy'])

checkpointer=ModelCheckpoint(save_best_only=True, filepath='nadam_model.h5')
## TODO: Train the model
nadam_hist = model.fit(X_train, y_train, batch_size=128, epochs=100, verbose=1, validation_split=0.2, callbacks=[checkpointer])

## TODO: Save the model as model.h5
model.save('nadam_model.h5')

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_13 (Conv2D)           (None, 96, 96, 8)         208       
_________________________________________________________________
activation_13 (Activation)   (None, 96, 96, 8)         0         
_________________________________________________________________
max_pooling2d_13 (MaxPooling (None, 24, 24, 8)         0         
_________________________________________________________________
conv2d_14 (Conv2D)           (None, 24, 24, 16)        1168      
_________________________________________________________________
activation_14 (Activation)   (None, 24, 24, 16)        0         
_________________________________________________________________
max_pooling2d_14 (MaxPooling (None, 6, 6, 16)          0         
_________________________________________________________________
flatten_7 (Flatten)          (None, 576)               0         
_________________________________________________________________
dense_7 (Dense)              (None, 30)                17310     
=================================================================
Total params: 18,686
Trainable params: 18,686
Non-trainable params: 0
_________________________________________________________________
Train on 1712 samples, validate on 428 samples
Epoch 1/100
1712/1712 [==============================] - 0s - loss: 0.0354 - acc: 0.4720 - val_loss: 0.0104 - val_acc: 0.6215
Epoch 2/100
1712/1712 [==============================] - 0s - loss: 0.0108 - acc: 0.6782 - val_loss: 0.0078 - val_acc: 0.6799
Epoch 3/100
1712/1712 [==============================] - 0s - loss: 0.0075 - acc: 0.7033 - val_loss: 0.0066 - val_acc: 0.6963
Epoch 4/100
1712/1712 [==============================] - 0s - loss: 0.0058 - acc: 0.7074 - val_loss: 0.0054 - val_acc: 0.6939
Epoch 5/100
1712/1712 [==============================] - 0s - loss: 0.0050 - acc: 0.7044 - val_loss: 0.0050 - val_acc: 0.6986
Epoch 6/100
1712/1712 [==============================] - 0s - loss: 0.0049 - acc: 0.7062 - val_loss: 0.0046 - val_acc: 0.6986
Epoch 7/100
1712/1712 [==============================] - 0s - loss: 0.0045 - acc: 0.7085 - val_loss: 0.0045 - val_acc: 0.6963
Epoch 8/100
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7085 - val_loss: 0.0046 - val_acc: 0.6986
Epoch 9/100
1712/1712 [==============================] - 0s - loss: 0.0045 - acc: 0.7074 - val_loss: 0.0047 - val_acc: 0.6986
Epoch 10/100
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7062 - val_loss: 0.0041 - val_acc: 0.7033
Epoch 11/100
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7097 - val_loss: 0.0037 - val_acc: 0.7033
Epoch 12/100
1712/1712 [==============================] - 0s - loss: 0.0040 - acc: 0.7109 - val_loss: 0.0051 - val_acc: 0.7033
Epoch 13/100
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7079 - val_loss: 0.0040 - val_acc: 0.7103
Epoch 14/100
1712/1712 [==============================] - 0s - loss: 0.0033 - acc: 0.7150 - val_loss: 0.0034 - val_acc: 0.7173
Epoch 15/100
1712/1712 [==============================] - 0s - loss: 0.0034 - acc: 0.7114 - val_loss: 0.0036 - val_acc: 0.7266
Epoch 16/100
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7109 - val_loss: 0.0036 - val_acc: 0.7243
Epoch 17/100
1712/1712 [==============================] - 0s - loss: 0.0031 - acc: 0.7126 - val_loss: 0.0028 - val_acc: 0.7266
Epoch 18/100
1712/1712 [==============================] - 0s - loss: 0.0031 - acc: 0.7173 - val_loss: 0.0033 - val_acc: 0.7313
Epoch 19/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7138 - val_loss: 0.0032 - val_acc: 0.7290
Epoch 20/100
1712/1712 [==============================] - 0s - loss: 0.0030 - acc: 0.7196 - val_loss: 0.0030 - val_acc: 0.7336
Epoch 21/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7202 - val_loss: 0.0030 - val_acc: 0.7243
Epoch 22/100
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7161 - val_loss: 0.0037 - val_acc: 0.7220
Epoch 23/100
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7202 - val_loss: 0.0031 - val_acc: 0.7290
Epoch 24/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7231 - val_loss: 0.0025 - val_acc: 0.7336
Epoch 25/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7161 - val_loss: 0.0034 - val_acc: 0.7313
Epoch 26/100
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7202 - val_loss: 0.0027 - val_acc: 0.7290
Epoch 27/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7190 - val_loss: 0.0032 - val_acc: 0.7313
Epoch 28/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7290 - val_loss: 0.0024 - val_acc: 0.7266
Epoch 29/100
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7266 - val_loss: 0.0030 - val_acc: 0.7407
Epoch 30/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7290 - val_loss: 0.0025 - val_acc: 0.7383
Epoch 31/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7225 - val_loss: 0.0024 - val_acc: 0.7430
Epoch 32/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7290 - val_loss: 0.0028 - val_acc: 0.7430
Epoch 33/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7243 - val_loss: 0.0024 - val_acc: 0.7523
Epoch 34/100
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7319 - val_loss: 0.0024 - val_acc: 0.7430
Epoch 35/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7261 - val_loss: 0.0026 - val_acc: 0.7430
Epoch 36/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7360 - val_loss: 0.0025 - val_acc: 0.7477
Epoch 37/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7331 - val_loss: 0.0029 - val_acc: 0.7383
Epoch 38/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7325 - val_loss: 0.0023 - val_acc: 0.7523
Epoch 39/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7348 - val_loss: 0.0022 - val_acc: 0.7523
Epoch 40/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7336 - val_loss: 0.0024 - val_acc: 0.7407
Epoch 41/100
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7354 - val_loss: 0.0025 - val_acc: 0.7477
Epoch 42/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7407 - val_loss: 0.0023 - val_acc: 0.7547
Epoch 43/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7436 - val_loss: 0.0023 - val_acc: 0.7570
Epoch 44/100
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7407 - val_loss: 0.0029 - val_acc: 0.7570
Epoch 45/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7389 - val_loss: 0.0022 - val_acc: 0.7547
Epoch 46/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7436 - val_loss: 0.0023 - val_acc: 0.7593
Epoch 47/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7471 - val_loss: 0.0024 - val_acc: 0.7523
Epoch 48/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7488 - val_loss: 0.0025 - val_acc: 0.7570
Epoch 49/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7459 - val_loss: 0.0023 - val_acc: 0.7570
Epoch 50/100
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7512 - val_loss: 0.0024 - val_acc: 0.7593
Epoch 51/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7500 - val_loss: 0.0021 - val_acc: 0.7617
Epoch 52/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7482 - val_loss: 0.0022 - val_acc: 0.7617
Epoch 53/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7512 - val_loss: 0.0024 - val_acc: 0.7593
Epoch 54/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7506 - val_loss: 0.0022 - val_acc: 0.7547
Epoch 55/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7535 - val_loss: 0.0022 - val_acc: 0.7570
Epoch 56/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7623 - val_loss: 0.0022 - val_acc: 0.7593
Epoch 57/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7535 - val_loss: 0.0020 - val_acc: 0.7500
Epoch 58/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7582 - val_loss: 0.0022 - val_acc: 0.7547
Epoch 59/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7570 - val_loss: 0.0021 - val_acc: 0.7593
Epoch 60/100
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7558 - val_loss: 0.0024 - val_acc: 0.7500
Epoch 61/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7629 - val_loss: 0.0022 - val_acc: 0.7593
Epoch 62/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7541 - val_loss: 0.0022 - val_acc: 0.7640
Epoch 63/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7599 - val_loss: 0.0022 - val_acc: 0.7687
Epoch 64/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7617 - val_loss: 0.0022 - val_acc: 0.7640
Epoch 65/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7576 - val_loss: 0.0023 - val_acc: 0.7640
Epoch 66/100
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7664 - val_loss: 0.0023 - val_acc: 0.7640
Epoch 67/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7629 - val_loss: 0.0019 - val_acc: 0.7664
Epoch 68/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7523 - val_loss: 0.0020 - val_acc: 0.7570
Epoch 69/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7588 - val_loss: 0.0019 - val_acc: 0.7734
Epoch 70/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7634 - val_loss: 0.0027 - val_acc: 0.7617
Epoch 71/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7588 - val_loss: 0.0022 - val_acc: 0.7570
Epoch 72/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7617 - val_loss: 0.0023 - val_acc: 0.7523
Epoch 73/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7623 - val_loss: 0.0019 - val_acc: 0.7664
Epoch 74/100
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7699 - val_loss: 0.0019 - val_acc: 0.7570
Epoch 75/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7640 - val_loss: 0.0021 - val_acc: 0.7710
Epoch 76/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7658 - val_loss: 0.0020 - val_acc: 0.7593
Epoch 77/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7687 - val_loss: 0.0020 - val_acc: 0.7617
Epoch 78/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7599 - val_loss: 0.0021 - val_acc: 0.7617
Epoch 79/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7687 - val_loss: 0.0019 - val_acc: 0.7640
Epoch 80/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7699 - val_loss: 0.0023 - val_acc: 0.7687
Epoch 81/100
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7664 - val_loss: 0.0020 - val_acc: 0.7617
Epoch 82/100
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7704 - val_loss: 0.0020 - val_acc: 0.7500
Epoch 83/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7734 - val_loss: 0.0020 - val_acc: 0.7640
Epoch 84/100
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7728 - val_loss: 0.0020 - val_acc: 0.7664
Epoch 85/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7658 - val_loss: 0.0020 - val_acc: 0.7804
Epoch 86/100
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7693 - val_loss: 0.0019 - val_acc: 0.7664
Epoch 87/100
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7734 - val_loss: 0.0019 - val_acc: 0.7757
Epoch 88/100
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7751 - val_loss: 0.0020 - val_acc: 0.7710
Epoch 89/100
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7775 - val_loss: 0.0021 - val_acc: 0.7664
Epoch 90/100
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7775 - val_loss: 0.0018 - val_acc: 0.7640
Epoch 91/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7775 - val_loss: 0.0021 - val_acc: 0.7687
Epoch 92/100
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7798 - val_loss: 0.0022 - val_acc: 0.7593
Epoch 93/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7734 - val_loss: 0.0021 - val_acc: 0.7710
Epoch 94/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7757 - val_loss: 0.0020 - val_acc: 0.7664
Epoch 95/100
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7804 - val_loss: 0.0019 - val_acc: 0.7757
Epoch 96/100
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7780 - val_loss: 0.0020 - val_acc: 0.7757
Epoch 97/100
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7786 - val_loss: 0.0021 - val_acc: 0.7804
Epoch 98/100
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7751 - val_loss: 0.0022 - val_acc: 0.7710
Epoch 99/100
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7786 - val_loss: 0.0023 - val_acc: 0.7687
Epoch 100/100
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7815 - val_loss: 0.0020 - val_acc: 0.7710

Step 7: Visualize the Loss and Test Predictions¶

(IMPLEMENTATION) Answer a few questions and visualize the loss¶

Question 1: Outline the steps you took to get to your final neural network architecture and your reasoning at each step.

Answer:

Since the input image has only 1 channel (gray), I have increased the depth of the input to 8 using the first convolutional layer. Also, since this is the first layer, I wanted to use a slightly bigger kernel of size 5 so as to increase the scope of capturing larger trends. This is followed by a relu acitvation and Max pooling layer of size 4. Relu activation helps the gradients to pass through more smoothly during backpropagation, and a larger kernel size of 4 is chosen so as to rapidly decrease the number of parameters that need to be fine-tuned. The second convolutional layer has a depth of 16 and a kernel size of 3. Typically, we increase the depth of the layers with every convolutional operation, which is why I've increased the depth from 8 to 16. A smaller kernel size of 3 is chosen since we want to narrow down the focus area with each successive convolutional steps. This is again followed by a relu activation and Max pooling layer. Finally, I've flattend the output and added a fully connected layer with 30 nodes on top, since we need to predict 30 points. No activation is added since we want a real number output. The overall architecture is built keeping in mind to restrict the number of trainable parameters below 20k for the purposes of easy trainability.

Question 2: Defend your choice of optimizer. Which optimizers did you test, and how did you determine which worked best?

Answer:

I've basically tried out all the optimizers. The comparision plot for the same is given below. The final optimizer is chosen purely based on its relative performance to keep the selection procedure as obective as possible. From the plot given below, it is clearly visible that Nadam optimizer is doing the best both in terms of training and validation accuracies. The training and validation losses are largely the same across the optimizers. Therefore, the Nadam optimizer is the one that got finally selected. Note, for the final run, the number of epochs have been doubled from (100 to 200) and dropout layer is added avoid overfitting.

Use the code cell below to plot the training and validation loss of your neural network. You may find this resource useful.

In [24]:

## TODO: Visualize the training and validation loss of your neural network
hists=[SGD_hist, rmsprop_hist, adagrad_hist, adam_hist, adadelta_hist, adamax_hist, nadam_hist]
names=['SGD', 'RMSprop', 'Adagrad', 'Adam', 'Adadelta', 'Adamax', 'Nadam']
colors=['b', 'g', 'r', 'y', 'k', 'm', 'c']

fig=plt.figure(figsize=(15,15))
ax1=fig.add_subplot(221)
ax2=fig.add_subplot(222)
ax3=fig.add_subplot(223)
ax4=fig.add_subplot(224)

ax1.set_title('Training Accuracy')
ax1.set_xlabel('Accuracy')
ax1.set_ylabel('Epoch')

ax2.set_title('Validation Accuracy')
ax2.set_xlabel('Accuracy')
ax2.set_ylabel('Epoch')

ax3.set_title('Training Loss')
ax3.set_xlabel('Loss')
ax3.set_ylabel('Epoch')

ax4.set_title('Validation Loss')
ax4.set_xlabel('Loss')
ax4.set_ylabel('Epoch')

for i, h in enumerate(hists):
    ax1.plot(h.history['acc'], c=colors[i], label=names[i])
    ax2.plot(h.history['val_acc'], c=colors[i], label=names[i])
    ax3.plot(h.history['loss'], c=colors[i], label=names[i])
    ax4.plot(h.history['val_loss'], c=colors[i], label=names[i])

fig.legend()
fig.tight_layout()
plt.show()

Chosen Optimizer - Nadam¶

In [19]:

#Nadam
from keras.models import Sequential
from keras.layers import Convolution2D, MaxPooling2D, Dropout, Activation
from keras.layers import Flatten, Dense
from keras.optimizers import SGD, RMSprop, Adagrad, Adadelta, Adam, Adamax, Nadam
from keras.callbacks import ModelCheckpoint
model = Sequential()

model.add(Convolution2D(8, kernel_size=5, padding='same',input_shape=X_train.shape[1:]))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Dropout(0.2))
model.add(Convolution2D(16, kernel_size=3, padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4,4), padding='same'))
model.add(Flatten())
model.add(Dense(30))
# Summarize the model
model.summary()

model.compile(optimizer='Nadam', loss='mean_squared_error', metrics=['accuracy'])

checkpointer=ModelCheckpoint(save_best_only=True, filepath='my_model.h5')
## TODO: Train the model
hist = model.fit(X_train, y_train, batch_size=128, epochs=200, verbose=1, validation_split=0.2, callbacks=[checkpointer])

## TODO: Save the model as model.h5
model.save('my_model.h5')

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_3 (Conv2D)            (None, 96, 96, 8)         208       
_________________________________________________________________
activation_3 (Activation)    (None, 96, 96, 8)         0         
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 24, 24, 8)         0         
_________________________________________________________________
dropout_2 (Dropout)          (None, 24, 24, 8)         0         
_________________________________________________________________
conv2d_4 (Conv2D)            (None, 24, 24, 16)        1168      
_________________________________________________________________
activation_4 (Activation)    (None, 24, 24, 16)        0         
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 6, 6, 16)          0         
_________________________________________________________________
flatten_2 (Flatten)          (None, 576)               0         
_________________________________________________________________
dense_2 (Dense)              (None, 30)                17310     
=================================================================
Total params: 18,686
Trainable params: 18,686
Non-trainable params: 0
_________________________________________________________________
Train on 1712 samples, validate on 428 samples
Epoch 1/200
1712/1712 [==============================] - 0s - loss: 0.0320 - acc: 0.3522 - val_loss: 0.0230 - val_acc: 0.4229
Epoch 2/200
1712/1712 [==============================] - 0s - loss: 0.0107 - acc: 0.5952 - val_loss: 0.0290 - val_acc: 0.5794
Epoch 3/200
1712/1712 [==============================] - 0s - loss: 0.0082 - acc: 0.6589 - val_loss: 0.0164 - val_acc: 0.6659
Epoch 4/200
1712/1712 [==============================] - 0s - loss: 0.0065 - acc: 0.6811 - val_loss: 0.0117 - val_acc: 0.6846
Epoch 5/200
1712/1712 [==============================] - 0s - loss: 0.0055 - acc: 0.6986 - val_loss: 0.0106 - val_acc: 0.6893
Epoch 6/200
1712/1712 [==============================] - 0s - loss: 0.0050 - acc: 0.6963 - val_loss: 0.0105 - val_acc: 0.6869
Epoch 7/200
1712/1712 [==============================] - 0s - loss: 0.0051 - acc: 0.7004 - val_loss: 0.0082 - val_acc: 0.6893
Epoch 8/200
1712/1712 [==============================] - 0s - loss: 0.0047 - acc: 0.7062 - val_loss: 0.0087 - val_acc: 0.6916
Epoch 9/200
1712/1712 [==============================] - 0s - loss: 0.0047 - acc: 0.7079 - val_loss: 0.0070 - val_acc: 0.6963
Epoch 10/200
1712/1712 [==============================] - 0s - loss: 0.0047 - acc: 0.7079 - val_loss: 0.0065 - val_acc: 0.6986
Epoch 11/200
1712/1712 [==============================] - 0s - loss: 0.0044 - acc: 0.7074 - val_loss: 0.0060 - val_acc: 0.7009
Epoch 12/200
1712/1712 [==============================] - 0s - loss: 0.0047 - acc: 0.7068 - val_loss: 0.0080 - val_acc: 0.70330.707
Epoch 13/200
1712/1712 [==============================] - 0s - loss: 0.0043 - acc: 0.7050 - val_loss: 0.0081 - val_acc: 0.7033
Epoch 14/200
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7091 - val_loss: 0.0047 - val_acc: 0.7079
Epoch 15/200
1712/1712 [==============================] - 0s - loss: 0.0045 - acc: 0.7085 - val_loss: 0.0050 - val_acc: 0.7033
Epoch 16/200
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7044 - val_loss: 0.0042 - val_acc: 0.7079
Epoch 17/200
1712/1712 [==============================] - 0s - loss: 0.0042 - acc: 0.7056 - val_loss: 0.0050 - val_acc: 0.7056
Epoch 18/200
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7027 - val_loss: 0.0052 - val_acc: 0.7103
Epoch 19/200
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7056 - val_loss: 0.0042 - val_acc: 0.7126
Epoch 20/200
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7091 - val_loss: 0.0043 - val_acc: 0.7220
Epoch 21/200
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7044 - val_loss: 0.0038 - val_acc: 0.7009
Epoch 22/200
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7033 - val_loss: 0.0057 - val_acc: 0.7009
Epoch 23/200
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7039 - val_loss: 0.0070 - val_acc: 0.7079
Epoch 24/200
1712/1712 [==============================] - 0s - loss: 0.0034 - acc: 0.7132 - val_loss: 0.0033 - val_acc: 0.7173
Epoch 25/200
1712/1712 [==============================] - 0s - loss: 0.0033 - acc: 0.7103 - val_loss: 0.0041 - val_acc: 0.7009
Epoch 26/200
1712/1712 [==============================] - 0s - loss: 0.0030 - acc: 0.7068 - val_loss: 0.0058 - val_acc: 0.7103
Epoch 27/200
1712/1712 [==============================] - 0s - loss: 0.0031 - acc: 0.7126 - val_loss: 0.0068 - val_acc: 0.6752
Epoch 28/200
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7068 - val_loss: 0.0052 - val_acc: 0.6986
Epoch 29/200
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7109 - val_loss: 0.0047 - val_acc: 0.7056
Epoch 30/200
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7033 - val_loss: 0.0046 - val_acc: 0.7150
Epoch 31/200
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7079 - val_loss: 0.0037 - val_acc: 0.7196
Epoch 32/200
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7114 - val_loss: 0.0039 - val_acc: 0.7033
Epoch 33/200
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7039 - val_loss: 0.0041 - val_acc: 0.7103
Epoch 34/200
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7068 - val_loss: 0.0042 - val_acc: 0.7079
Epoch 35/200
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7173 - val_loss: 0.0072 - val_acc: 0.7103
Epoch 36/200
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7062 - val_loss: 0.0045 - val_acc: 0.7009
Epoch 37/200
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7015 - val_loss: 0.0050 - val_acc: 0.6986
Epoch 38/200
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7109 - val_loss: 0.0040 - val_acc: 0.7150
Epoch 39/200
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7062 - val_loss: 0.0036 - val_acc: 0.7103
Epoch 40/200
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7155 - val_loss: 0.0043 - val_acc: 0.7173
Epoch 41/200
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7155 - val_loss: 0.0044 - val_acc: 0.7103
Epoch 42/200
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7132 - val_loss: 0.0025 - val_acc: 0.70790.71
Epoch 43/200
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7097 - val_loss: 0.0025 - val_acc: 0.7220
Epoch 44/200
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7132 - val_loss: 0.0032 - val_acc: 0.7126
Epoch 45/200
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7179 - val_loss: 0.0031 - val_acc: 0.7196
Epoch 46/200
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7120 - val_loss: 0.0027 - val_acc: 0.7126
Epoch 47/200
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7173 - val_loss: 0.0025 - val_acc: 0.7220
Epoch 48/200
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7144 - val_loss: 0.0027 - val_acc: 0.7173
Epoch 49/200
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7132 - val_loss: 0.0028 - val_acc: 0.71960.712
Epoch 50/200
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7185 - val_loss: 0.0024 - val_acc: 0.7150
Epoch 51/200
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7173 - val_loss: 0.0036 - val_acc: 0.7243
Epoch 52/200
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7202 - val_loss: 0.0028 - val_acc: 0.7220
Epoch 53/200
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7190 - val_loss: 0.0036 - val_acc: 0.7290
Epoch 54/200
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7150 - val_loss: 0.0039 - val_acc: 0.7243
Epoch 55/200
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7208 - val_loss: 0.0023 - val_acc: 0.7079
Epoch 56/200
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7255 - val_loss: 0.0031 - val_acc: 0.7173
Epoch 57/200
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7220 - val_loss: 0.0028 - val_acc: 0.7173
Epoch 58/200
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7231 - val_loss: 0.0026 - val_acc: 0.7243
Epoch 59/200
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7173 - val_loss: 0.0039 - val_acc: 0.7266
Epoch 60/200
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7150 - val_loss: 0.0036 - val_acc: 0.7290
Epoch 61/200
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7179 - val_loss: 0.0032 - val_acc: 0.7313
Epoch 62/200
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7150 - val_loss: 0.0025 - val_acc: 0.7336
Epoch 63/200
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7185 - val_loss: 0.0024 - val_acc: 0.7196
Epoch 64/200
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7342 - val_loss: 0.0023 - val_acc: 0.7290
Epoch 65/200
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7208 - val_loss: 0.0042 - val_acc: 0.7336
Epoch 66/200
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7301 - val_loss: 0.0024 - val_acc: 0.7290
Epoch 67/200
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7202 - val_loss: 0.0036 - val_acc: 0.7313
Epoch 68/200
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7103 - val_loss: 0.0022 - val_acc: 0.7290
Epoch 69/200
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7366 - val_loss: 0.0024 - val_acc: 0.7360
Epoch 70/200
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7348 - val_loss: 0.0029 - val_acc: 0.7336
Epoch 71/200
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7266 - val_loss: 0.0028 - val_acc: 0.7383
Epoch 72/200
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7255 - val_loss: 0.0026 - val_acc: 0.7360
Epoch 73/200
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7284 - val_loss: 0.0027 - val_acc: 0.7453
Epoch 74/200
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7348 - val_loss: 0.0031 - val_acc: 0.7290
Epoch 75/200
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7336 - val_loss: 0.0021 - val_acc: 0.7383
Epoch 76/200
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7307 - val_loss: 0.0024 - val_acc: 0.7453
Epoch 77/200
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7261 - val_loss: 0.0026 - val_acc: 0.7196
Epoch 78/200
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7342 - val_loss: 0.0026 - val_acc: 0.7079
Epoch 79/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7296 - val_loss: 0.0021 - val_acc: 0.7336
Epoch 80/200
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7331 - val_loss: 0.0020 - val_acc: 0.7336
Epoch 81/200
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7266 - val_loss: 0.0022 - val_acc: 0.7336
Epoch 82/200
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7313 - val_loss: 0.0020 - val_acc: 0.7383
Epoch 83/200
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7418 - val_loss: 0.0026 - val_acc: 0.7243
Epoch 84/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7342 - val_loss: 0.0022 - val_acc: 0.7430
Epoch 85/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7389 - val_loss: 0.0031 - val_acc: 0.7383
Epoch 86/200
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7371 - val_loss: 0.0023 - val_acc: 0.7360
Epoch 87/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7395 - val_loss: 0.0023 - val_acc: 0.7407
Epoch 88/200
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7354 - val_loss: 0.0020 - val_acc: 0.7360
Epoch 89/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7284 - val_loss: 0.0019 - val_acc: 0.7336
Epoch 90/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7430 - val_loss: 0.0029 - val_acc: 0.7523
Epoch 91/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7453 - val_loss: 0.0023 - val_acc: 0.7266
Epoch 92/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7354 - val_loss: 0.0022 - val_acc: 0.7477
Epoch 93/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7272 - val_loss: 0.0024 - val_acc: 0.7430
Epoch 94/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7342 - val_loss: 0.0029 - val_acc: 0.7430
Epoch 95/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7389 - val_loss: 0.0019 - val_acc: 0.7266
Epoch 96/200
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7383 - val_loss: 0.0020 - val_acc: 0.7336
Epoch 97/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7301 - val_loss: 0.0020 - val_acc: 0.7196
Epoch 98/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7319 - val_loss: 0.0018 - val_acc: 0.7430
Epoch 99/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7401 - val_loss: 0.0020 - val_acc: 0.7453
Epoch 100/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7488 - val_loss: 0.0024 - val_acc: 0.7430
Epoch 101/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7430 - val_loss: 0.0020 - val_acc: 0.7430
Epoch 102/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7383 - val_loss: 0.0018 - val_acc: 0.7453
Epoch 103/200
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.7401 - val_loss: 0.0019 - val_acc: 0.7570
Epoch 104/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7336 - val_loss: 0.0022 - val_acc: 0.7453
Epoch 105/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7377 - val_loss: 0.0027 - val_acc: 0.7547
Epoch 106/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7418 - val_loss: 0.0024 - val_acc: 0.7687
Epoch 107/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7477 - val_loss: 0.0019 - val_acc: 0.7453
Epoch 108/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7383 - val_loss: 0.0019 - val_acc: 0.7477
Epoch 109/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7447 - val_loss: 0.0022 - val_acc: 0.7570
Epoch 110/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7436 - val_loss: 0.0024 - val_acc: 0.7570
Epoch 111/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7366 - val_loss: 0.0024 - val_acc: 0.7617
Epoch 112/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7471 - val_loss: 0.0018 - val_acc: 0.7547
Epoch 113/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7412 - val_loss: 0.0019 - val_acc: 0.7664
Epoch 114/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7453 - val_loss: 0.0022 - val_acc: 0.7523
Epoch 115/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7471 - val_loss: 0.0019 - val_acc: 0.7453
Epoch 116/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7442 - val_loss: 0.0018 - val_acc: 0.7547
Epoch 117/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7605 - val_loss: 0.0021 - val_acc: 0.7664
Epoch 118/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7535 - val_loss: 0.0018 - val_acc: 0.7477
Epoch 119/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7465 - val_loss: 0.0017 - val_acc: 0.7523
Epoch 120/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7453 - val_loss: 0.0018 - val_acc: 0.7523
Epoch 121/200
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7477 - val_loss: 0.0019 - val_acc: 0.7640
Epoch 122/200
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7442 - val_loss: 0.0021 - val_acc: 0.7547
Epoch 123/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7488 - val_loss: 0.0027 - val_acc: 0.7430
Epoch 124/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7523 - val_loss: 0.0019 - val_acc: 0.7500
Epoch 125/200
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7459 - val_loss: 0.0020 - val_acc: 0.7734
Epoch 126/200
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7453 - val_loss: 0.0018 - val_acc: 0.7500
Epoch 127/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7412 - val_loss: 0.0017 - val_acc: 0.7593
Epoch 128/200
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7488 - val_loss: 0.0022 - val_acc: 0.7617
Epoch 129/200
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7553 - val_loss: 0.0021 - val_acc: 0.7547
Epoch 130/200
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7442 - val_loss: 0.0027 - val_acc: 0.7617
Epoch 131/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7523 - val_loss: 0.0023 - val_acc: 0.7710
Epoch 132/200
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7453 - val_loss: 0.0021 - val_acc: 0.7804
Epoch 133/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7447 - val_loss: 0.0020 - val_acc: 0.7734
Epoch 134/200
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7401 - val_loss: 0.0019 - val_acc: 0.7640
Epoch 135/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7570 - val_loss: 0.0020 - val_acc: 0.7710
Epoch 136/200
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7482 - val_loss: 0.0018 - val_acc: 0.7710
Epoch 137/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7506 - val_loss: 0.0017 - val_acc: 0.79440.
Epoch 138/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7623 - val_loss: 0.0017 - val_acc: 0.7687
Epoch 139/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7512 - val_loss: 0.0016 - val_acc: 0.7710
Epoch 140/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7488 - val_loss: 0.0017 - val_acc: 0.7617
Epoch 141/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7629 - val_loss: 0.0018 - val_acc: 0.7710
Epoch 142/200
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7523 - val_loss: 0.0017 - val_acc: 0.7827
Epoch 143/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7611 - val_loss: 0.0017 - val_acc: 0.7850
Epoch 144/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7500 - val_loss: 0.0024 - val_acc: 0.7593
Epoch 145/200
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7442 - val_loss: 0.0022 - val_acc: 0.7500
Epoch 146/200
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.7465 - val_loss: 0.0018 - val_acc: 0.7664
Epoch 147/200
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7535 - val_loss: 0.0018 - val_acc: 0.7687
Epoch 148/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7588 - val_loss: 0.0016 - val_acc: 0.7664
Epoch 149/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7588 - val_loss: 0.0018 - val_acc: 0.7500
Epoch 150/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7453 - val_loss: 0.0025 - val_acc: 0.7360
Epoch 151/200
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7699 - val_loss: 0.0020 - val_acc: 0.7780
Epoch 152/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7611 - val_loss: 0.0018 - val_acc: 0.7780
Epoch 153/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7623 - val_loss: 0.0022 - val_acc: 0.7664
Epoch 154/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7634 - val_loss: 0.0018 - val_acc: 0.7640
Epoch 155/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7535 - val_loss: 0.0021 - val_acc: 0.7570
Epoch 156/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7693 - val_loss: 0.0016 - val_acc: 0.7523
Epoch 157/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7664 - val_loss: 0.0019 - val_acc: 0.7827
Epoch 158/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7664 - val_loss: 0.0024 - val_acc: 0.7827
Epoch 159/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7523 - val_loss: 0.0017 - val_acc: 0.7757
Epoch 160/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7617 - val_loss: 0.0016 - val_acc: 0.7687
Epoch 161/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7564 - val_loss: 0.0019 - val_acc: 0.7477
Epoch 162/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7728 - val_loss: 0.0019 - val_acc: 0.76170.771
Epoch 163/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7623 - val_loss: 0.0019 - val_acc: 0.7710
Epoch 164/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7716 - val_loss: 0.0018 - val_acc: 0.7804
Epoch 165/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7605 - val_loss: 0.0019 - val_acc: 0.7687
Epoch 166/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7541 - val_loss: 0.0018 - val_acc: 0.7617
Epoch 167/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7786 - val_loss: 0.0019 - val_acc: 0.7710
Epoch 168/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7611 - val_loss: 0.0017 - val_acc: 0.7360
Epoch 169/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7605 - val_loss: 0.0016 - val_acc: 0.7757
Epoch 170/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7728 - val_loss: 0.0019 - val_acc: 0.7850
Epoch 171/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7652 - val_loss: 0.0021 - val_acc: 0.76870.76
Epoch 172/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7704 - val_loss: 0.0015 - val_acc: 0.7780
Epoch 173/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7728 - val_loss: 0.0015 - val_acc: 0.7897
Epoch 174/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7722 - val_loss: 0.0015 - val_acc: 0.7827
Epoch 175/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7699 - val_loss: 0.0016 - val_acc: 0.7757
Epoch 176/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7745 - val_loss: 0.0016 - val_acc: 0.7780
Epoch 177/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7734 - val_loss: 0.0016 - val_acc: 0.7874
Epoch 178/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7652 - val_loss: 0.0016 - val_acc: 0.7523
Epoch 179/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7716 - val_loss: 0.0015 - val_acc: 0.7804
Epoch 180/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7757 - val_loss: 0.0015 - val_acc: 0.7804
Epoch 181/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7734 - val_loss: 0.0015 - val_acc: 0.7757
Epoch 182/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7757 - val_loss: 0.0020 - val_acc: 0.7850
Epoch 183/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7710 - val_loss: 0.0016 - val_acc: 0.7710
Epoch 184/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7710 - val_loss: 0.0017 - val_acc: 0.7827
Epoch 185/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7658 - val_loss: 0.0018 - val_acc: 0.7757
Epoch 186/200
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7728 - val_loss: 0.0019 - val_acc: 0.7897
Epoch 187/200
1712/1712 [==============================] - ETA: 0s - loss: 0.0012 - acc: 0.784 - 0s - loss: 0.0013 - acc: 0.7839 - val_loss: 0.0015 - val_acc: 0.7710
Epoch 188/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7780 - val_loss: 0.0017 - val_acc: 0.7850
Epoch 189/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7763 - val_loss: 0.0021 - val_acc: 0.7780
Epoch 190/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7734 - val_loss: 0.0018 - val_acc: 0.7617
Epoch 191/200
1712/1712 [==============================] - 0s - loss: 0.0012 - acc: 0.7775 - val_loss: 0.0016 - val_acc: 0.7570
Epoch 192/200
1712/1712 [==============================] - 0s - loss: 0.0012 - acc: 0.7739 - val_loss: 0.0019 - val_acc: 0.7804
Epoch 193/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7792 - val_loss: 0.0015 - val_acc: 0.7850
Epoch 194/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7815 - val_loss: 0.0016 - val_acc: 0.7687
Epoch 195/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7769 - val_loss: 0.0015 - val_acc: 0.7874
Epoch 196/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7728 - val_loss: 0.0017 - val_acc: 0.7757
Epoch 197/200
1712/1712 [==============================] - 0s - loss: 0.0012 - acc: 0.7710 - val_loss: 0.0015 - val_acc: 0.7874
Epoch 198/200
1712/1712 [==============================] - 0s - loss: 0.0012 - acc: 0.7780 - val_loss: 0.0018 - val_acc: 0.7687
Epoch 199/200
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7874 - val_loss: 0.0017 - val_acc: 0.7664
Epoch 200/200
1712/1712 [==============================] - 0s - loss: 0.0012 - acc: 0.7757 - val_loss: 0.0016 - val_acc: 0.7991

In [43]:

fig=plt.figure(figsize=(15,7))
ax1=fig.add_subplot(121)
ax2=fig.add_subplot(122)

ax1.plot(hist.history['acc'], label='Training', c='b')
ax1.plot(hist.history['val_acc'], label='Testing', c='g')
ax1.set_xlabel('Epochs')
ax1.set_ylabel('Accuracy')
ax1.legend()

ax2.plot(hist.history['loss'], label='Training', c='b')
ax2.plot(hist.history['val_loss'], label='Testing', c='g')
ax2.set_xlabel('Epochs')
ax2.set_ylabel('Loss')
ax2.legend()

fig.suptitle('Nadam Optimizer', fontsize=15)
plt.show()

Question 3: Do you notice any evidence of overfitting or underfitting in the above plot? If so, what steps have you taken to improve your model? Note that slight overfitting or underfitting will not hurt your chances of a successful submission, as long as you have attempted some solutions towards improving your model (such as regularization, dropout, increased/decreased number of layers, etc).

Answer:

From the above plot, it is apparent that there is not much variation between training and validation accuracies/losses. This is an indication that the model is not overfitting. Besides, the architecture used is not overly complicated and the added dropout layer also helps in avoiding overfitting, if any. Also, the loss function is more or less saturated, and also reached the mentioned optimal loss of 0.0015. Therefore, we can safely conclude that the model is not heavily underfitted either.

Visualize a Subset of the Test Predictions¶

Execute the code cell below to visualize your model's predicted keypoints on a subset of the testing images.

In [44]:

y_test = model.predict(X_test)
fig = plt.figure(figsize=(20,20))
fig.subplots_adjust(left=0, right=1, bottom=0, top=1, hspace=0.05, wspace=0.05)
for i in range(9):
    ax = fig.add_subplot(3, 3, i + 1, xticks=[], yticks=[])
    plot_data(X_test[i], y_test[i], ax)

Step 8: Complete the pipeline¶

With the work you did in Sections 1 and 2 of this notebook, along with your freshly trained facial keypoint detector, you can now complete the full pipeline. That is given a color image containing a person or persons you can now

Detect the faces in this image automatically using OpenCV
Predict the facial keypoints in each face detected in the image
Paint predicted keypoints on each face detected

In this Subsection you will do just this!

(IMPLEMENTATION) Facial Keypoints Detector¶

Use the OpenCV face detection functionality you built in previous Sections to expand the functionality of your keypoints detector to color images with arbitrary size. Your function should perform the following steps

Accept a color image.
Convert the image to grayscale.
Detect and crop the face contained in the image.
Locate the facial keypoints in the cropped image.
Overlay the facial keypoints in the original (color, uncropped) image.

Note: step 4 can be the trickiest because remember your convolutional network is only trained to detect facial keypoints in $96 \times 96$ grayscale images where each pixel was normalized to lie in the interval $[0,1]$, and remember that each facial keypoint was normalized during training to the interval $[-1,1]$. This means - practically speaking - to paint detected keypoints onto a test face you need to perform this same pre-processing to your candidate face - that is after detecting it you should resize it to $96 \times 96$ and normalize its values before feeding it into your facial keypoint detector. To be shown correctly on the original image the output keypoints from your detector then need to be shifted and re-normalized from the interval $[-1,1]$ to the width and height of your detected face.

When complete you should be able to produce example images like the one below

In [20]:

# Load in color image for face detection
image = cv2.imread('images/obamas4.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)


# plot our image
fig = plt.figure(figsize = (9,9))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.set_title('image copy')
ax1.imshow(image)

Out[20]:

<matplotlib.image.AxesImage at 0x7fdc43feeb70>

In [21]:

### TODO: Use the face detection code we saw in Section 1 with your trained conv-net 
## TODO : Paint the predicted keypoints on the test image

gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)

faces = face_cascade.detectMultiScale(gray, 2 ,3)
print('number of faces detected: ', len(faces))

image_copy = np.copy(image)
landmarks_list=[]
for (x, y, w, h) in faces:
    cv2.rectangle(image_copy, (x, y), (x+w, y+h), (255, 0, 0), 3)
    crop_image = gray[y:y+h, x:x+h]
    resize_image = cv2.resize(crop_image, (96, 96))
    normalized_image = (255-resize_image)/255
    y_test = model.predict(normalized_image.reshape(-1, 96, 96, 1))
    x_landmarks = y_test[:,0::2]
    x_landmarks.reshape(15)
    y_landmarks = y_test[:,1::2]
    y_landmarks.reshape(15)
    x_landmarks = (x_landmarks*w/2)+(w/2)+x
    y_landmarks = (y_landmarks*h/2)+(h/2)+y
    landmarks_list.append((x_landmarks, y_landmarks))
    
fig=plt.figure(figsize=(8,8))
ax1=fig.add_subplot(111)
ax1.imshow(image_copy)
ax1.set_xticks([])
ax1.set_yticks([])
for landmark in landmarks_list:
    ax1.scatter(landmark[0], landmark[1], marker='o', c='c', s=10)
plt.show()

number of faces detected:  2

(Optional) Further Directions - add a filter using facial keypoints to your laptop camera¶

Now you can add facial keypoint detection to your laptop camera - as illustrated in the gif below.

The next Python cell contains the basic laptop video camera function used in the previous optional video exercises. Combine it with the functionality you developed for keypoint detection and marking in the previous exercise and you should be good to go!

In [ ]:

import cv2
import time 
from keras.models import load_model
def laptop_camera_go():
    # Create instance of video capturer
    cv2.namedWindow("face detection activated")
    vc = cv2.VideoCapture(0)

    # Try to get the first frame
    if vc.isOpened(): 
        rval, frame = vc.read()
    else:
        rval = False
    
    # keep video stream open
    while rval:
        # plot image from camera with detections marked
        cv2.imshow("face detection activated", frame)
        
        # exit functionality - press any key to exit laptop video
        key = cv2.waitKey(20)
        if key > 0: # exit by pressing any key
            # destroy windows
            cv2.destroyAllWindows()
            
            # hack from stack overflow for making sure window closes on osx --> https://stackoverflow.com/questions/6116564/destroywindow-does-not-close-window-on-mac-using-python-and-opencv
            for i in range (1,5):
                cv2.waitKey(1)
            return
        
        # read next frame
        time.sleep(0.05)             # control framerate for computation - default 20 frames per sec
        rval, frame = vc.read()

In [ ]:

# Run your keypoint face painter
laptop_camera_go()

(Optional) Further Directions - add a filter using facial keypoints¶

Using your freshly minted facial keypoint detector pipeline you can now do things like add fun filters to a person's face automatically. In this optional exercise you can play around with adding sunglasses automatically to each individual's face in an image as shown in a demonstration image below.

To produce this effect an image of a pair of sunglasses shown in the Python cell below.

In [22]:

# Load in sunglasses image - note the usage of the special option
# cv2.IMREAD_UNCHANGED, this option is used because the sunglasses 
# image has a 4th channel that allows us to control how transparent each pixel in the image is
sunglasses = cv2.imread("images/sunglasses_4.png", cv2.IMREAD_UNCHANGED)

# Plot the image
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.imshow(sunglasses)
ax1.axis('off');

This image is placed over each individual's face using the detected eye points to determine the location of the sunglasses, and eyebrow points to determine the size that the sunglasses should be for each person (one could also use the nose point to determine this).

Notice that this image actually has 4 channels, not just 3.

In [50]:

# Print out the shape of the sunglasses image
print ('The sunglasses image has shape: ' + str(np.shape(sunglasses)))

The sunglasses image has shape: (1123, 3064, 4)

It has the usual red, blue, and green channels any color image has, with the 4th channel representing the transparency level of each pixel in the image. Here's how the transparency channel works: the lower the value, the more transparent the pixel will become. The lower bound (completely transparent) is zero here, so any pixels set to 0 will not be seen.

This is how we can place this image of sunglasses on someone's face and still see the area around of their face where the sunglasses lie - because these pixels in the sunglasses image have been made completely transparent.

Lets check out the alpha channel of our sunglasses image in the next Python cell. Note because many of the pixels near the boundary are transparent we'll need to explicitly print out non-zero values if we want to see them.

In [23]:

# Print out the sunglasses transparency (alpha) channel
alpha_channel = sunglasses[:,:,3]
print ('the alpha channel here looks like')
print (alpha_channel)

# Just to double check that there are indeed non-zero values
# Let's find and print out every value greater than zero
values = np.where(alpha_channel != 0)
print ('\n the non-zero values of the alpha channel look like')
print (values)

the alpha channel here looks like
[[0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 ..., 
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]]

 the non-zero values of the alpha channel look like
(array([  17,   17,   17, ..., 1109, 1109, 1109]), array([ 687,  688,  689, ..., 2376, 2377, 2378]))

This means that when we place this sunglasses image on top of another image, we can use the transparency channel as a filter to tell us which pixels to overlay on a new image (only the non-transparent ones with values greater than zero).

One last thing: it's helpful to understand which keypoint belongs to the eyes, mouth, etc. So, in the image below, we also display the index of each facial keypoint directly on the image so that you can tell which keypoints are for the eyes, eyebrows, etc.

With this information, you're well on your way to completing this filtering task! See if you can place the sunglasses automatically on the individuals in the image loaded in / shown in the next Python cell.

In [82]:

# Load in color image for face detection
image = cv2.imread('images/obamas4.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)


# Plot the image
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.set_title('Original Image')
ax1.imshow(image)

Out[82]:

<matplotlib.image.AxesImage at 0x7fd89b6e7cf8>

In [69]:

## (Optional) TODO: Use the face detection code we saw in Section 1 with your trained conv-net to put
## sunglasses on the individuals in our test image

### TODO: Use the face detection code we saw in Section 1 with your trained conv-net 
## TODO : Paint the predicted keypoints on the test image

gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)

faces = face_cascade.detectMultiScale(gray, 2 ,3)
print('number of faces detected: ', len(faces))

image_copy = np.copy(image)
landmarks_list=[]
for (x, y, w, h) in faces:
    color_image_crop=image_copy[y:y+h, x:x+w]
    #cv2.rectangle(image_copy, (x, y), (x+w, y+h), (255, 0, 0), 3)
    crop_image = gray[y:y+h, x:x+h]
    resize_image = cv2.resize(crop_image, (96, 96))
    normalized_image = (255-resize_image)/255
    y_test = model.predict(normalized_image.reshape(-1, 96, 96, 1))
    x_landmarks = y_test[:,0::2]
    x_landmarks.reshape(15)
    y_landmarks = y_test[:,1::2]
    y_landmarks.reshape(15)
    x_landmarks = (x_landmarks*w/2)+(w/2)+x
    y_landmarks = (y_landmarks*h/2)+(h/2)+y
    landmarks_list.append((x_landmarks, y_landmarks))
    
    x_eyebrows=1.1*((x_landmarks[0][7]-x_landmarks[0][9]))-10
    y_eyebrows = 2*((y_landmarks[0][4]-y_landmarks[0][8]))+10
    
    sunglasses_resized = cv2.resize(sunglasses[:,:,], (int(x_eyebrows), int(y_eyebrows)))
    
    orig_mask = sunglasses_resized[:,:,3]
    
    x_range1=int(x_landmarks[0][9]-x)
    x_range2=int(x_landmarks[0][9]-x+sunglasses_resized.shape[1])
    y_range1=int(y_landmarks[0][8]-y)
    y_range2=int(y_landmarks[0][8]-y+sunglasses_resized.shape[0])
    
    mask=sunglasses_resized[:,:,3]
    inv_mask=cv2.bitwise_not(mask)
    
    img1=cv2.bitwise_and(color_image_crop[y_range1:y_range2, x_range1:x_range2], 
                            color_image_crop[y_range1:y_range2, x_range1:x_range2], mask=inv_mask)
    
    img2=cv2.bitwise_and(sunglasses_resized[:,:,:3], sunglasses_resized[:,:,:3], mask=mask)
    final_image = cv2.add(img1, img2)
    
    image_copy[y_range1+y:y_range2+y, x_range1+x:x_range2+x]=final_image
    
    plt.imshow(final_image)
    plt.xticks([])
    plt.yticks([])
    plt.show()
    
fig=plt.figure(figsize=(8,8))
ax1=fig.add_subplot(111)
ax1.imshow(image_copy)
ax1.set_xticks([])
ax1.set_yticks([])
plt.show()

number of faces detected:  2

(Optional) Further Directions - add a filter using facial keypoints to your laptop camera¶

Now you can add the sunglasses filter to your laptop camera - as illustrated in the gif below.

The next Python cell contains the basic laptop video camera function used in the previous optional video exercises. Combine it with the functionality you developed for adding sunglasses to someone's face in the previous optional exercise and you should be good to go!

In [ ]:

import cv2
import time 
from keras.models import load_model
import numpy as np

def laptop_camera_go():
    # Create instance of video capturer
    cv2.namedWindow("face detection activated")
    vc = cv2.VideoCapture(0)

    # try to get the first frame
    if vc.isOpened(): 
        rval, frame = vc.read()
    else:
        rval = False
    
    # Keep video stream open
    while rval:
        # Plot image from camera with detections marked
        cv2.imshow("face detection activated", frame)
        
        # Exit functionality - press any key to exit laptop video
        key = cv2.waitKey(20)
        if key > 0: # exit by pressing any key
            # Destroy windows 
            cv2.destroyAllWindows()
            
            for i in range (1,5):
                cv2.waitKey(1)
            return
        
        # Read next frame
        time.sleep(0.05)             # control framerate for computation - default 20 frames per sec
        rval, frame = vc.read()

In [ ]:

# Load facial landmark detector model
model = load_model('my_model.h5')

# Run sunglasses painter
laptop_camera_go()

links

social